add GElan proposed by YOLOv9

config/__init__.py

@@ -7,12 +7,13 @@ from .yolov5_af_config import build_yolov5af_config
 from .yolov6_config    import build_yolov6_config
 from .yolov7_af_config import build_yolov7af_config
 from .yolov8_config    import build_yolov8_config
+from .gelan_config     import build_gelan_config
 from .rtdetr_config    import build_rtdetr_config
 
 def build_config(args):
     print('==============================')
     print('Model: {} ...'.format(args.model.upper()))
-    # YOLO series
+    # ----------- YOLO series -----------
     if   'yolov1' in args.model:
         cfg = build_yolov1_config(args)
     elif 'yolov2' in args.model:
@@ -29,7 +30,9 @@ def build_config(args):
         cfg = build_yolov7af_config(args)
     elif 'yolov8' in args.model:
         cfg = build_yolov8_config(args)
-    # RT-DETR
+    elif 'gelan' in args.model:
+        cfg = build_gelan_config(args)
+    # ----------- RT-DETR -----------
     elif 'rtdetr' in args.model:
         cfg = build_rtdetr_config(args)
     else:

config/gelan_config.py

@@ -0,0 +1,142 @@
+# Gelan (proposed by yolov9) config
+
+
+def build_gelan_config(args):
+    if   args.model == 'gelan_c':
+        return GElanCConfig()
+    else:
+        raise NotImplementedError("No config for model: {}".format(args.model))
+    
+# GELAN-Base config
+class GElanBaseConfig(object):
+    def __init__(self) -> None:
+        # ---------------- Model config ----------------
+        self.reg_max  = 16
+        self.out_stride = [8, 16, 32]
+        self.max_stride = 32
+        self.num_levels = 3
+        ## Backbone
+        self.bk_act = 'silu'
+        self.bk_norm = 'BN'
+        self.bk_depthwise = False
+        self.bk_down_pooling = True
+        self.backbone_feats = {
+            "c1": [64],
+            "c2": [128, [128, 64],  256],
+            "c3": [256, [256, 128], 512],
+            "c4": [512, [512, 256], 512],
+            "c5": [512, [512, 256], 512],
+        }
+        self.backbone_depth = 1
+        ## Neck
+        self.neck           = 'spp_elan'
+        self.neck_act       = 'silu'
+        self.neck_norm      = 'BN'
+        self.spp_pooling_size  = 5
+        self.spp_inter_dim     = 256
+        self.spp_out_dim       = 512
+        ## FPN
+        self.fpn      = 'gelan_pafpn'
+        self.fpn_act  = 'silu'
+        self.fpn_norm = 'BN'
+        self.fpn_depthwise = False
+        self.fpn_down_pooling = True
+        self.fpn_depth    = 1
+        self.fpn_feats_td = {
+            "p4": [[512, 256], 512],
+            "p3": [[256, 128], 256],
+        }
+        self.fpn_feats_bu = {
+            "p4": [[512, 256], 512],
+            "p5": [[512, 256], 512],
+        }
+        ## Head
+        self.head      = 'gelan_head'
+        self.head_act  = 'silu'
+        self.head_norm = 'BN'
+        self.head_depthwise = False
+        self.num_cls_head   = 2
+        self.num_reg_head   = 2
+
+        # ---------------- Post-process config ----------------
+        ## Post process
+        self.val_topk = 1000
+        self.val_conf_thresh = 0.001
+        self.val_nms_thresh  = 0.7
+        self.test_topk = 100
+        self.test_conf_thresh = 0.2
+        self.test_nms_thresh  = 0.5
+
+        # ---------------- Assignment config ----------------
+        ## Matcher
+        self.tal_topk_candidates = 10
+        self.tal_alpha = 0.5
+        self.tal_beta  = 6.0
+        ## Loss weight
+        self.loss_cls = 0.5
+        self.loss_box = 7.5
+        self.loss_dfl = 1.5
+
+        # ---------------- ModelEMA config ----------------
+        self.use_ema = True
+        self.ema_decay = 0.9998
+        self.ema_tau   = 2000
+
+        # ---------------- Optimizer config ----------------
+        self.trainer      = 'yolo'
+        self.optimizer    = 'adamw'
+        self.per_image_lr = 0.001 / 64
+        self.base_lr      = None      # base_lr = per_image_lr * batch_size
+        self.min_lr_ratio = 0.01      # min_lr  = base_lr * min_lr_ratio
+        self.momentum     = 0.9
+        self.weight_decay = 0.05
+        self.clip_max_norm   = -1.
+        self.warmup_bias_lr  = 0.1
+        self.warmup_momentum = 0.8
+
+        # ---------------- Lr Scheduler config ----------------
+        self.warmup_epoch = 3
+        self.lr_scheduler = "cosine"
+        self.max_epoch    = 500
+        self.eval_epoch   = 10
+        self.no_aug_epoch = 20
+
+        # ---------------- Data process config ----------------
+        self.aug_type = 'yolo'
+        self.box_format = 'xyxy'
+        self.normalize_coords = False
+        self.mosaic_prob = 0.0
+        self.mixup_prob  = 0.0
+        self.copy_paste  = 0.0           # approximated by the YOLOX's mixup
+        self.multi_scale = [0.5, 1.25]   # multi scale: [img_size * 0.5, img_size * 1.25]
+        ## Pixel mean & std
+        self.pixel_mean = [0., 0., 0.]
+        self.pixel_std  = [255., 255., 255.]
+        ## Transforms
+        self.train_img_size = 640
+        self.test_img_size  = 640
+        self.use_ablu = True
+        self.affine_params = {
+            'degrees': 0.0,
+            'translate': 0.2,
+            'scale': [0.1, 2.0],
+            'shear': 0.0,
+            'perspective': 0.0,
+            'hsv_h': 0.015,
+            'hsv_s': 0.7,
+            'hsv_v': 0.4,
+        }
+
+    def print_config(self):
+        config_dict = {key: value for key, value in self.__dict__.items() if not key.startswith('__')}
+        for k, v in config_dict.items():
+            print("{} : {}".format(k, v))
+
+# GELAN-C
+class GElanCConfig(GElanBaseConfig):
+    def __init__(self) -> None:
+        super().__init__()
+        # ---------------- Data process config ----------------
+        self.mosaic_prob = 1.0
+        self.mixup_prob  = 0.1
+        self.copy_paste  = 0.5

config/yolov6_config.py

@@ -54,8 +54,8 @@ class Yolov6BaseConfig(object):
 
         # ---------------- Assignment config ----------------
         ## Matcher
-        self.tal_topk_candidates = 10
-        self.tal_alpha = 0.5
+        self.tal_topk_candidates = 13
+        self.tal_alpha = 1.0
         self.tal_beta  = 6.0
         ## Loss weight
         self.loss_cls = 1.0

models/__init__.py

@@ -10,6 +10,7 @@ from .yolov5_af.build import build_yolov5af
 from .yolov6.build    import build_yolov6
 from .yolov7_af.build import build_yolov7af
 from .yolov8.build    import build_yolov8
+from .gelan.build     import build_gelan
 from .rtdetr.build    import build_rtdetr
 
 # build object detector
@@ -39,6 +40,9 @@ def build_model(args, cfg, is_val=False):
     ## YOLOv8
     elif 'yolov8' in args.model:
         model, criterion = build_yolov8(cfg, is_val)
+    ## GElan
+    elif 'gelan' in args.model:
+        model, criterion = build_gelan(cfg, is_val)
     ## RT-DETR
     elif 'rtdetr' in args.model:
         model, criterion = build_rtdetr(cfg, is_val)

models/gelan/build.py

@@ -0,0 +1,24 @@
+import torch.nn as nn
+
+from .loss import SetCriterion
+from .gelan import GElan
+
+
+# build object detector
+def build_gelan(cfg, is_val=False):
+    # -------------- Build YOLO --------------
+    model = GElan(cfg, is_val, deploy=False)
+
+    # -------------- Initialize YOLO --------------
+    for m in model.modules():
+        if isinstance(m, nn.BatchNorm2d):
+            m.eps = 1e-3
+            m.momentum = 0.03    
+            
+    # -------------- Build criterion --------------
+    criterion = None
+    if is_val:
+        # build criterion for training
+        criterion = SetCriterion(cfg)
+        
+    return model, criterion

models/gelan/gelan.py

@@ -0,0 +1,165 @@
+# --------------- Torch components ---------------
+import torch
+import torch.nn as nn
+
+# --------------- Model components ---------------
+from .gelan_backbone import GElanBackbone
+from .gelan_neck     import SPPElan
+from .gelan_pafpn    import GElanPaFPN
+from .gelan_head     import GElanDetHead
+from .gelan_pred     import GElanPredLayer
+
+# --------------- External components ---------------
+from utils.misc import multiclass_nms
+
+
+# G-ELAN proposed by YOLOv9
+class GElan(nn.Module):
+    def __init__(self,
+                 cfg,
+                 is_val = False,
+                 deploy = False,
+                 ) -> None:
+        super(GElan, self).__init__()
+        # ---------------------- Basic setting ----------------------
+        self.cfg = cfg
+        self.deploy = deploy
+        self.num_classes = cfg.num_classes
+        ## Post-process parameters
+        self.topk_candidates = cfg.val_topk        if is_val else cfg.test_topk
+        self.conf_thresh     = cfg.val_conf_thresh if is_val else cfg.test_conf_thresh
+        self.nms_thresh      = cfg.val_nms_thresh  if is_val else cfg.test_nms_thresh
+        self.no_multi_labels = False if is_val else True
+        
+        # ---------------------- Network Parameters ----------------------
+        ## Backbone
+        self.backbone = GElanBackbone(cfg)
+        self.neck     = SPPElan(cfg, self.backbone.feat_dims[-1])
+        self.backbone.feat_dims[-1] = self.neck.out_dim
+        ## PaFPN
+        self.fpn      = GElanPaFPN(cfg, self.backbone.feat_dims)
+        ## Detection head
+        self.head     = GElanDetHead(cfg, self.fpn.out_dims)
+        self.pred     = GElanPredLayer(cfg, self.head.cls_head_dim, self.head.reg_head_dim)
+
+    def switch_to_deploy(self,):
+        for m in self.modules():
+            if hasattr(m, "fuse_convs"):
+                m.fuse_convs()
+
+    def post_process(self, cls_preds, box_preds):
+        """
+        Input:
+            cls_preds: List[torch.Tensor] -> [[B, M, C], ...], B=1
+            box_preds: List[torch.Tensor] -> [[B, M, 4], ...], B=1
+        Output:
+            bboxes: np.array -> [N, 4]
+            scores: np.array -> [N,]
+            labels: np.array -> [N,]
+        """
+        all_scores = []
+        all_labels = []
+        all_bboxes = []
+        
+        for cls_pred_i, box_pred_i in zip(cls_preds, box_preds):
+            cls_pred_i = cls_pred_i[0]
+            box_pred_i = box_pred_i[0]
+            if self.no_multi_labels:
+                # [M,]
+                scores, labels = torch.max(cls_pred_i.sigmoid(), dim=1)
+
+                # Keep top k top scoring indices only.
+                num_topk = min(self.topk_candidates, box_pred_i.size(0))
+
+                # topk candidates
+                predicted_prob, topk_idxs = scores.sort(descending=True)
+                topk_scores = predicted_prob[:num_topk]
+                topk_idxs = topk_idxs[:num_topk]
+
+                # filter out the proposals with low confidence score
+                keep_idxs = topk_scores > self.conf_thresh
+                scores = topk_scores[keep_idxs]
+                topk_idxs = topk_idxs[keep_idxs]
+
+                labels = labels[topk_idxs]
+                bboxes = box_pred_i[topk_idxs]
+            else:
+                # [M, C] -> [MC,]
+                scores_i = cls_pred_i.sigmoid().flatten()
+
+                # Keep top k top scoring indices only.
+                num_topk = min(self.topk_candidates, box_pred_i.size(0))
+
+                # torch.sort is actually faster than .topk (at least on GPUs)
+                predicted_prob, topk_idxs = scores_i.sort(descending=True)
+                topk_scores = predicted_prob[:num_topk]
+                topk_idxs = topk_idxs[:num_topk]
+
+                # filter out the proposals with low confidence score
+                keep_idxs = topk_scores > self.conf_thresh
+                scores = topk_scores[keep_idxs]
+                topk_idxs = topk_idxs[keep_idxs]
+
+                anchor_idxs = torch.div(topk_idxs, self.num_classes, rounding_mode='floor')
+                labels = topk_idxs % self.num_classes
+
+                bboxes = box_pred_i[anchor_idxs]
+
+            all_scores.append(scores)
+            all_labels.append(labels)
+            all_bboxes.append(bboxes)
+
+        scores = torch.cat(all_scores, dim=0)
+        labels = torch.cat(all_labels, dim=0)
+        bboxes = torch.cat(all_bboxes, dim=0)
+
+        # to cpu & numpy
+        scores = scores.cpu().numpy()
+        labels = labels.cpu().numpy()
+        bboxes = bboxes.cpu().numpy()
+
+        # nms
+        scores, labels, bboxes = multiclass_nms(
+            scores, labels, bboxes, self.nms_thresh, self.num_classes)
+
+        return bboxes, scores, labels
+    
+    def forward(self, x):
+        # ---------------- Backbone ----------------
+        pyramid_feats = self.backbone(x)
+        # ---------------- Neck: SPP ----------------
+        pyramid_feats[-1] = self.neck(pyramid_feats[-1])
+
+        # ---------------- Neck: PaFPN ----------------
+        pyramid_feats = self.fpn(pyramid_feats)
+
+        # ---------------- Heads ----------------
+        cls_feats, reg_feats = self.head(pyramid_feats)
+
+        # ---------------- Preds ----------------
+        outputs = self.pred(cls_feats, reg_feats)
+        outputs['image_size'] = [x.shape[2], x.shape[3]]
+
+        if not self.training:
+            all_cls_preds = outputs['pred_cls']
+            all_box_preds = outputs['pred_box']
+
+            if self.deploy:
+                cls_preds = torch.cat(all_cls_preds, dim=1)[0]
+                box_preds = torch.cat(all_box_preds, dim=1)[0]
+                scores = cls_preds.sigmoid()
+                bboxes = box_preds
+                # [n_anchors_all, 4 + C]
+                outputs = torch.cat([bboxes, scores], dim=-1)
+
+            else:
+                # post process
+                bboxes, scores, labels = self.post_process(all_cls_preds, all_box_preds)
+                outputs = {
+                    "scores": scores,
+                    "labels": labels,
+                    "bboxes": bboxes
+                }
+        
+        return outputs
+    

models/gelan/gelan_backbone.py

@@ -0,0 +1,149 @@
+import torch
+import torch.nn as nn
+
+try:
+    from .gelan_basic import BasicConv, RepGElanLayer, ADown
+except:
+    from  gelan_basic import BasicConv, RepGElanLayer, ADown
+
+
+# ---------------------------- Basic functions ----------------------------
+class GElanBackbone(nn.Module):
+    def __init__(self, cfg):
+        super(GElanBackbone, self).__init__()
+        # ------------------ Basic setting ------------------
+        self.feat_dims = [cfg.backbone_feats["c1"][-1],  # 64
+                          cfg.backbone_feats["c2"][-1],  # 128
+                          cfg.backbone_feats["c3"][-1],  # 256
+                          cfg.backbone_feats["c4"][-1],  # 512
+                          cfg.backbone_feats["c5"][-1],  # 512
+                          ]
+        
+        # ------------------ Network setting ------------------
+        ## P1/2
+        self.layer_1 = BasicConv(3, cfg.backbone_feats["c1"][0],
+                                 kernel_size=3, padding=1, stride=2,
+                                 act_type=cfg.bk_act, norm_type=cfg.bk_norm, depthwise=cfg.bk_depthwise)
+        # P2/4
+        self.layer_2 = nn.Sequential(
+            BasicConv(cfg.backbone_feats["c1"][0], cfg.backbone_feats["c2"][0],
+                      kernel_size=3, padding=1, stride=2,
+                      act_type=cfg.bk_act, norm_type=cfg.bk_norm, depthwise=cfg.bk_depthwise),
+            RepGElanLayer(in_dim     = cfg.backbone_feats["c2"][0],
+                          inter_dims = cfg.backbone_feats["c2"][1],
+                          out_dim    = cfg.backbone_feats["c2"][2],
+                          num_blocks = cfg.backbone_depth,
+                          shortcut   = True,
+                          act_type   = cfg.bk_act,
+                          norm_type  = cfg.bk_norm,
+                          depthwise  = cfg.bk_depthwise)
+        )
+        # P3/8
+        self.layer_3 = nn.Sequential(
+            ADown(cfg.backbone_feats["c2"][2], cfg.backbone_feats["c3"][0],
+                  act_type=cfg.bk_act, norm_type=cfg.bk_norm, depthwise=cfg.bk_depthwise, use_pooling=cfg.bk_down_pooling),
+            RepGElanLayer(in_dim     = cfg.backbone_feats["c3"][0],
+                          inter_dims = cfg.backbone_feats["c3"][1],
+                          out_dim    = cfg.backbone_feats["c3"][2],
+                          num_blocks = cfg.backbone_depth,
+                          shortcut   = True,
+                          act_type   = cfg.bk_act,
+                          norm_type  = cfg.bk_norm,
+                          depthwise  = cfg.bk_depthwise)
+        )
+        # P4/16
+        self.layer_4 = nn.Sequential(
+            ADown(cfg.backbone_feats["c3"][2], cfg.backbone_feats["c4"][0],
+                  act_type=cfg.bk_act, norm_type=cfg.bk_norm, depthwise=cfg.bk_depthwise, use_pooling=cfg.bk_down_pooling),
+            RepGElanLayer(in_dim     = cfg.backbone_feats["c4"][0],
+                          inter_dims = cfg.backbone_feats["c4"][1],
+                          out_dim    = cfg.backbone_feats["c4"][2],
+                          num_blocks = cfg.backbone_depth,
+                          shortcut   = True,
+                          act_type   = cfg.bk_act,
+                          norm_type  = cfg.bk_norm,
+                          depthwise  = cfg.bk_depthwise)
+        )
+        # P5/32
+        self.layer_5 = nn.Sequential(
+            ADown(cfg.backbone_feats["c4"][2], cfg.backbone_feats["c5"][0],
+                  act_type=cfg.bk_act, norm_type=cfg.bk_norm, depthwise=cfg.bk_depthwise, use_pooling=cfg.bk_down_pooling),
+            RepGElanLayer(in_dim     = cfg.backbone_feats["c5"][0],
+                          inter_dims = cfg.backbone_feats["c5"][1],
+                          out_dim    = cfg.backbone_feats["c5"][2],
+                          num_blocks = cfg.backbone_depth,
+                          shortcut   = True,
+                          act_type   = cfg.bk_act,
+                          norm_type  = cfg.bk_norm,
+                          depthwise  = cfg.bk_depthwise)
+        )
+
+        # Initialize all layers
+        self.init_weights()
+        
+    def init_weights(self):
+        """Initialize the parameters."""
+        for m in self.modules():
+            if isinstance(m, torch.nn.Conv2d):
+                # In order to be consistent with the source code,
+                # reset the Conv2d initialization parameters
+                m.reset_parameters()
+
+    def forward(self, x):
+        c1 = self.layer_1(x)
+        c2 = self.layer_2(c1)
+        c3 = self.layer_3(c2)
+        c4 = self.layer_4(c3)
+        c5 = self.layer_5(c4)
+        outputs = [c3, c4, c5]
+
+        return outputs
+
+
+# ---------------------------- Functions ----------------------------
+## build Yolo's Backbone
+def build_backbone(cfg): 
+    # model
+    backbone = GElanBackbone(cfg)
+        
+    return backbone
+
+
+if __name__ == '__main__':
+    import time
+    from thop import profile
+    base_config = {
+        "bk_act": "silu",
+        "bk_norm": "BN"
+    }
+    class BaseConfig(object):
+        def __init__(self) -> None:
+            self.bk_act = 'silu'
+            self.bk_norm = 'BN'
+            self.bk_depthwise = False
+            self.backbone_feats = {
+                "c1": [64],
+                "c2": [128, [128, 64], 256],
+                "c3": [256, [256, 128], 512],
+                "c4": [512, [512, 256], 512],
+                "c5": [512, [512, 256], 512],
+            }
+            self.backbone_depth = 1
+
+    cfg = BaseConfig()
+    model = build_backbone(cfg)
+    x = torch.randn(1, 3, 640, 640)
+    t0 = time.time()
+    outputs = model(x)
+    t1 = time.time()
+    print('Time: ', t1 - t0)
+    for out in outputs:
+        print(out.shape)
+
+    x = torch.randn(1, 3, 640, 640)
+    print('==============================')
+    flops, params = profile(model, inputs=(x, ), verbose=False)
+    print('==============================')
+    print('GFLOPs : {:.2f}'.format(flops / 1e9 * 2))
+    print('Params : {:.2f} M'.format(params / 1e6))
+    

models/gelan/gelan_basic.py

@@ -0,0 +1,320 @@
+import numpy as np
+import torch
+import torch.nn as nn
+from typing import List
+
+
+# --------------------- Basic modules ---------------------
+def get_conv2d(c1, c2, k, p, s, d, g, bias=False):
+    conv = nn.Conv2d(c1, c2, k, stride=s, padding=p, dilation=d, groups=g, bias=bias)
+
+    return conv
+
+def get_activation(act_type=None):
+    if act_type == 'relu':
+        return nn.ReLU(inplace=True)
+    elif act_type == 'lrelu':
+        return nn.LeakyReLU(0.1, inplace=True)
+    elif act_type == 'mish':
+        return nn.Mish(inplace=True)
+    elif act_type == 'silu':
+        return nn.SiLU(inplace=True)
+    elif act_type is None:
+        return nn.Identity()
+    else:
+        raise NotImplementedError
+        
+def get_norm(norm_type, dim):
+    if norm_type == 'BN':
+        return nn.BatchNorm2d(dim)
+    elif norm_type == 'GN':
+        return nn.GroupNorm(num_groups=32, num_channels=dim)
+    elif norm_type is None:
+        return nn.Identity()
+    else:
+        raise NotImplementedError
+
+class BasicConv(nn.Module):
+    def __init__(self, 
+                 in_dim,                   # in channels
+                 out_dim,                  # out channels 
+                 kernel_size=1,            # kernel size 
+                 padding=0,                # padding
+                 stride=1,                 # padding
+                 dilation=1,               # dilation
+                 group=1,                  # group
+                 act_type  :str = 'lrelu', # activation
+                 norm_type :str = 'BN',    # normalization
+                 depthwise :bool = False
+                ):
+        super(BasicConv, self).__init__()
+        self.depthwise = depthwise
+        if not depthwise:
+            self.conv = get_conv2d(in_dim, out_dim, k=kernel_size, p=padding, s=stride, d=dilation, g=group)
+            self.norm = get_norm(norm_type, out_dim)
+        else:
+            self.conv1 = get_conv2d(in_dim, in_dim, k=kernel_size, p=padding, s=stride, d=dilation, g=in_dim)
+            self.norm1 = get_norm(norm_type, in_dim)
+            self.conv2 = get_conv2d(in_dim, out_dim, k=1, p=0, s=1, d=1, g=1)
+            self.norm2 = get_norm(norm_type, out_dim)
+        self.act  = get_activation(act_type)
+
+    def forward(self, x):
+        if not self.depthwise:
+            return self.act(self.norm(self.conv(x)))
+        else:
+            # Depthwise conv
+            x = self.norm1(self.conv1(x))
+            # Pointwise conv
+            x = self.norm2(self.conv2(x))
+            return x
+
+
+# --------------------- GELAN modules (from yolov9) ---------------------
+class ADown(nn.Module):
+    def __init__(self, in_dim, out_dim, act_type="silu", norm_type="BN", depthwise=False, use_pooling=True):
+        super().__init__()
+        inter_dim = out_dim // 2
+        self.use_pooling = use_pooling
+        if use_pooling:
+            self.conv_layer_1 = BasicConv(in_dim // 2, inter_dim,
+                                        kernel_size=3, padding=1, stride=2,
+                                        act_type=act_type, norm_type=norm_type, depthwise=depthwise)
+            self.conv_layer_2 = BasicConv(in_dim // 2, inter_dim, kernel_size=1,
+                                        act_type=act_type, norm_type=norm_type, depthwise=depthwise)
+        else:
+            self.conv_layer = BasicConv(in_dim, out_dim, kernel_size=3, padding=1, stride=2,
+                                        act_type=act_type, norm_type=norm_type, depthwise=depthwise)
+    def forward(self, x):
+        if self.use_pooling:
+            x = torch.nn.functional.avg_pool2d(x, 2, 1, 0, False, True)
+            x1,x2 = x.chunk(2, 1)
+            x1 = self.conv_layer_1(x1)
+            x2 = torch.nn.functional.max_pool2d(x2, 3, 2, 1)
+            x2 = self.conv_layer_2(x2)
+
+            return torch.cat((x1, x2), 1)
+        else:
+            return self.conv_layer(x)
+
+class RepConvN(nn.Module):
+    """RepConv is a basic rep-style block, including training and deploy status
+    This code is based on https://github.com/DingXiaoH/RepVGG/blob/main/repvgg.py
+    """
+    def __init__(self, in_dim, out_dim, k=3, s=1, p=1, g=1, act_type='silu', norm_type='BN', depthwise=False):
+        super().__init__()
+        assert k == 3 and p == 1
+        self.g = g
+        self.in_dim = in_dim
+        self.out_dim = out_dim
+        self.act = get_activation(act_type)
+
+        self.bn = None
+        self.conv1 = BasicConv(in_dim, out_dim,
+                               kernel_size=k, padding=p, stride=s, group=g,
+                               act_type=None, norm_type=norm_type, depthwise=depthwise)
+        self.conv2 = BasicConv(in_dim, out_dim,
+                               kernel_size=1, padding=(p - k // 2), stride=s, group=g,
+                               act_type=None, norm_type=norm_type, depthwise=depthwise)
+
+    def forward(self, x):
+        """Forward process"""
+        if hasattr(self, 'conv'):
+            return self.forward_fuse(x)
+        else:
+            id_out = 0 if self.bn is None else self.bn(x)
+            return self.act(self.conv1(x) + self.conv2(x) + id_out)
+
+    def forward_fuse(self, x):
+        """Forward process"""
+        return self.act(self.conv(x))
+
+    def get_equivalent_kernel_bias(self):
+        kernel3x3, bias3x3 = self._fuse_bn_tensor(self.conv1)
+        kernel1x1, bias1x1 = self._fuse_bn_tensor(self.conv2)
+        kernelid, biasid = self._fuse_bn_tensor(self.bn)
+        return kernel3x3 + self._pad_1x1_to_3x3_tensor(kernel1x1) + kernelid, bias3x3 + bias1x1 + biasid
+
+    def _avg_to_3x3_tensor(self, avgp):
+        channels = self.in_dim
+        groups = self.g
+        kernel_size = avgp.kernel_size
+        input_dim = channels // groups
+        k = torch.zeros((channels, input_dim, kernel_size, kernel_size))
+        k[np.arange(channels), np.tile(np.arange(input_dim), groups), :, :] = 1.0 / kernel_size ** 2
+        return k
+
+    def _pad_1x1_to_3x3_tensor(self, kernel1x1):
+        if kernel1x1 is None:
+            return 0
+        else:
+            return torch.nn.functional.pad(kernel1x1, [1, 1, 1, 1])
+
+    def _fuse_bn_tensor(self, branch):
+        if branch is None:
+            return 0, 0
+        if isinstance(branch, BasicConv):
+            kernel       = branch.conv.weight
+            running_mean = branch.norm.running_mean
+            running_var  = branch.norm.running_var
+            gamma        = branch.norm.weight
+            beta         = branch.norm.bias
+            eps          = branch.norm.eps
+        elif isinstance(branch, nn.BatchNorm2d):
+            if not hasattr(self, 'id_tensor'):
+                input_dim = self.in_dim // self.g
+                kernel_value = np.zeros((self.in_dim, input_dim, 3, 3), dtype=np.float32)
+                for i in range(self.in_dim):
+                    kernel_value[i, i % input_dim, 1, 1] = 1
+                self.id_tensor = torch.from_numpy(kernel_value).to(branch.weight.device)
+            kernel       = self.id_tensor
+            running_mean = branch.running_mean
+            running_var  = branch.running_var
+            gamma        = branch.weight
+            beta         = branch.bias
+            eps          = branch.eps
+        std = (running_var + eps).sqrt()
+        t = (gamma / std).reshape(-1, 1, 1, 1)
+        return kernel * t, beta - running_mean * gamma / std
+
+    def fuse_convs(self):
+        if hasattr(self, 'conv'):
+            return
+        kernel, bias = self.get_equivalent_kernel_bias()
+        self.conv = nn.Conv2d(in_channels  = self.conv1.conv.in_channels,
+                              out_channels = self.conv1.conv.out_channels,
+                              kernel_size  = self.conv1.conv.kernel_size,
+                              stride       = self.conv1.conv.stride,
+                              padding      = self.conv1.conv.padding,
+                              dilation     = self.conv1.conv.dilation,
+                              groups       = self.conv1.conv.groups,
+                              bias         = True).requires_grad_(False)
+        self.conv.weight.data = kernel
+        self.conv.bias.data = bias
+        for para in self.parameters():
+            para.detach_()
+        self.__delattr__('conv1')
+        self.__delattr__('conv2')
+        if hasattr(self, 'nm'):
+            self.__delattr__('nm')
+        if hasattr(self, 'bn'):
+            self.__delattr__('bn')
+        if hasattr(self, 'id_tensor'):
+            self.__delattr__('id_tensor')
+
+class RepNBottleneck(nn.Module):
+    def __init__(self,
+                 in_dim,
+                 out_dim,
+                 shortcut=True,
+                 kernel_size=(3, 3),
+                 expansion=0.5,
+                 act_type='silu',
+                 norm_type='BN',
+                 depthwise=False
+                 ):
+        super().__init__()
+        inter_dim = round(out_dim * expansion)
+        self.conv_layer_1 = RepConvN(in_dim, inter_dim, kernel_size[0], p=kernel_size[0]//2, s=1, act_type=act_type, norm_type=norm_type)
+        self.conv_layer_2 = BasicConv(inter_dim, out_dim, kernel_size[1], padding=kernel_size[1]//2, stride=1, act_type=act_type, norm_type=norm_type)
+        self.add = shortcut and in_dim == out_dim
+
+    def forward(self, x):
+        h = self.conv_layer_2(self.conv_layer_1(x))
+        return x + h if self.add else h
+
+class RepNCSP(nn.Module):
+    def __init__(self,
+                 in_dim,
+                 out_dim,
+                 num_blocks=1,
+                 shortcut=True,
+                 expansion=0.5,
+                 act_type='silu',
+                 norm_type='BN',
+                 depthwise=False
+                 ):
+        super().__init__()
+        inter_dim = int(out_dim * expansion)
+        self.conv_layer_1 = BasicConv(in_dim, inter_dim, kernel_size=1, act_type=act_type, norm_type=norm_type)
+        self.conv_layer_2 = BasicConv(in_dim, inter_dim, kernel_size=1, act_type=act_type, norm_type=norm_type)
+        self.conv_layer_3 = BasicConv(2 * inter_dim, out_dim, kernel_size=1)
+        self.module       = nn.Sequential(*(RepNBottleneck(inter_dim,
+                                                           inter_dim,
+                                                           kernel_size = [3, 3],
+                                                           shortcut    = shortcut,
+                                                           expansion   = 1.0,
+                                                           act_type    = act_type,
+                                                           norm_type   = norm_type,
+                                                           depthwise   = depthwise)
+                                                           for _ in range(num_blocks)))
+
+    def forward(self, x):
+        x1 = self.conv_layer_1(x)
+        x2 = self.module(self.conv_layer_2(x))
+
+        return self.conv_layer_3(torch.cat([x1, x2], dim=1))
+
+class RepGElanLayer(nn.Module):
+    """YOLOv9's GELAN module"""
+    def __init__(self,
+                 in_dim     :int,
+                 inter_dims :List,
+                 out_dim    :int,
+                 num_blocks :int   = 1,
+                 shortcut   :bool  = False,
+                 act_type   :str   = 'silu',
+                 norm_type  :str   = 'BN',
+                 depthwise  :bool  = False,
+                 ) -> None:
+        super(RepGElanLayer, self).__init__()
+        # ----------- Basic parameters -----------
+        self.in_dim = in_dim
+        self.inter_dims = inter_dims
+        self.out_dim = out_dim
+
+        # ----------- Network parameters -----------
+        self.conv_layer_1  = BasicConv(in_dim, inter_dims[0], kernel_size=1, act_type=act_type, norm_type=norm_type)
+        self.elan_module_1 = nn.Sequential(
+             RepNCSP(inter_dims[0]//2,
+                     inter_dims[1],
+                     num_blocks  = num_blocks,
+                     shortcut    = shortcut,
+                     expansion   = 0.5,
+                     act_type    = act_type,
+                     norm_type   = norm_type,
+                     depthwise   = depthwise),
+            BasicConv(inter_dims[1], inter_dims[1],
+                      kernel_size=3, padding=1,
+                      act_type=act_type, norm_type=norm_type, depthwise=depthwise)
+        )
+        self.elan_module_2 = nn.Sequential(
+             RepNCSP(inter_dims[1],
+                     inter_dims[1],
+                     num_blocks  = num_blocks,
+                     shortcut    = shortcut,
+                     expansion   = 0.5,
+                     act_type    = act_type,
+                     norm_type   = norm_type,
+                     depthwise   = depthwise),
+            BasicConv(inter_dims[1], inter_dims[1],
+                      kernel_size=3, padding=1,
+                      act_type=act_type, norm_type=norm_type, depthwise=depthwise)
+        )
+        self.conv_layer_2 = BasicConv(inter_dims[0] + 2*self.inter_dims[1], out_dim, kernel_size=1, act_type=act_type, norm_type=norm_type)
+
+
+    def forward(self, x):
+        # Input proj
+        x1, x2 = torch.chunk(self.conv_layer_1(x), 2, dim=1)
+        out = list([x1, x2])
+
+        # ELAN module
+        out.append(self.elan_module_1(out[-1]))
+        out.append(self.elan_module_2(out[-1]))
+
+        # Output proj
+        out = self.conv_layer_2(torch.cat(out, dim=1))
+
+        return out
+    

models/gelan/gelan_head.py

@@ -0,0 +1,125 @@
+import torch
+import torch.nn as nn
+
+from .gelan_basic import BasicConv
+
+
+# Single-level Head
+class SingleLevelHead(nn.Module):
+    def __init__(self,
+                 in_dim       :int  = 256,
+                 cls_head_dim :int  = 256,
+                 reg_head_dim :int  = 256,
+                 num_cls_head :int  = 2,
+                 num_reg_head :int  = 2,
+                 act_type     :str  = "silu",
+                 norm_type    :str  = "BN",
+                 depthwise    :bool = False):
+        super().__init__()
+        # --------- Basic Parameters ----------
+        self.in_dim = in_dim
+        self.num_cls_head = num_cls_head
+        self.num_reg_head = num_reg_head
+        self.act_type = act_type
+        self.norm_type = norm_type
+        self.depthwise = depthwise
+        
+        # --------- Network Parameters ----------
+        ## cls head
+        cls_feats = []
+        self.cls_head_dim = cls_head_dim
+        for i in range(num_cls_head):
+            if i == 0:
+                cls_feats.append(
+                    BasicConv(in_dim, self.cls_head_dim,
+                              kernel_size=3, padding=1, stride=1, 
+                              act_type=act_type,
+                              norm_type=norm_type,
+                              depthwise=depthwise)
+                              )
+            else:
+                cls_feats.append(
+                    BasicConv(self.cls_head_dim, self.cls_head_dim,
+                              kernel_size=3, padding=1, stride=1, 
+                              act_type=act_type,
+                              norm_type=norm_type,
+                              depthwise=depthwise)
+                              )
+        ## reg head
+        reg_feats = []
+        self.reg_head_dim = reg_head_dim
+        for i in range(num_reg_head):
+            if i == 0:
+                reg_feats.append(
+                    BasicConv(in_dim, self.reg_head_dim,
+                              kernel_size=3, padding=1, stride=1, 
+                              act_type=act_type,
+                              norm_type=norm_type,
+                              depthwise=depthwise)
+                              )
+            else:
+                reg_feats.append(
+                    BasicConv(self.reg_head_dim, self.reg_head_dim,
+                              kernel_size=3, padding=1, stride=1, group=4,
+                              act_type=act_type,
+                              norm_type=norm_type,
+                              depthwise=depthwise)
+                              )
+        self.cls_feats = nn.Sequential(*cls_feats)
+        self.reg_feats = nn.Sequential(*reg_feats)
+
+        self.init_weights()
+        
+    def init_weights(self):
+        """Initialize the parameters."""
+        for m in self.modules():
+            if isinstance(m, torch.nn.Conv2d):
+                # In order to be consistent with the source code,
+                # reset the Conv2d initialization parameters
+                m.reset_parameters()
+
+    def forward(self, x):
+        """
+            in_feats: (Tensor) [B, C, H, W]
+        """
+        cls_feats = self.cls_feats(x)
+        reg_feats = self.reg_feats(x)
+
+        return cls_feats, reg_feats
+    
+# Multi-level Head
+class GElanDetHead(nn.Module):
+    def __init__(self, cfg, in_dims):
+        super().__init__()
+        ## ----------- Network Parameters -----------
+        self.multi_level_heads = nn.ModuleList(
+            [SingleLevelHead(in_dim       = in_dims[level],
+                             cls_head_dim = max(in_dims[0], min(cfg.num_classes * 2, 128)),
+                             reg_head_dim = max(in_dims[0]//4, 16, 4*cfg.reg_max),
+                             num_cls_head = cfg.num_cls_head,
+                             num_reg_head = cfg.num_reg_head,
+                             act_type     = cfg.head_act,
+                             norm_type    = cfg.head_norm,
+                             depthwise    = cfg.head_depthwise)
+                             for level in range(cfg.num_levels)
+                             ])
+        # --------- Basic Parameters ----------
+        self.in_dims = in_dims
+        self.cls_head_dim = self.multi_level_heads[0].cls_head_dim
+        self.reg_head_dim = self.multi_level_heads[0].reg_head_dim
+
+
+    def forward(self, feats):
+        """
+            feats: List[(Tensor)] [[B, C, H, W], ...]
+        """
+        cls_feats = []
+        reg_feats = []
+        for feat, head in zip(feats, self.multi_level_heads):
+            # ---------------- Pred ----------------
+            cls_feat, reg_feat = head(feat)
+
+            cls_feats.append(cls_feat)
+            reg_feats.append(reg_feat)
+
+        return cls_feats, reg_feats

models/gelan/gelan_neck.py

@@ -0,0 +1,54 @@
+import torch
+import torch.nn as nn
+
+from .gelan_basic import BasicConv
+
+
+# SPPF (from yolov5)
+class SPPF(nn.Module):
+    """
+        This code referenced to https://github.com/ultralytics/yolov5
+    """
+    def __init__(self, cfg, in_dim, out_dim):
+        super().__init__()
+        ## ----------- Basic Parameters -----------
+        inter_dim = round(in_dim * cfg.neck_expand_ratio)
+        self.out_dim = out_dim
+        ## ----------- Network Parameters -----------
+        self.cv1 = BasicConv(in_dim, inter_dim,
+                             kernel_size=1, padding=0, stride=1,
+                             act_type=cfg.neck_act, norm_type=cfg.neck_norm)
+        self.cv2 = BasicConv(inter_dim * 4, out_dim,
+                             kernel_size=1, padding=0, stride=1,
+                             act_type=cfg.neck_act, norm_type=cfg.neck_norm)
+        self.m = nn.MaxPool2d(kernel_size=cfg.spp_pooling_size,
+                              stride=1,
+                              padding=cfg.spp_pooling_size // 2)
+
+    def forward(self, x):
+        x = self.cv1(x)
+        y1 = self.m(x)
+        y2 = self.m(y1)
+
+        return self.cv2(torch.cat((x, y1, y2, self.m(y2)), 1))
+
+# SPP-ELAN (from yolov9)
+class SPPElan(nn.Module):
+    def __init__(self, cfg, in_dim):
+        """SPPElan looks like the SPPF."""
+        super().__init__()
+        ## ----------- Basic Parameters -----------
+        self.in_dim = in_dim
+        self.inter_dim = cfg.spp_inter_dim
+        self.out_dim   = cfg.spp_out_dim
+        ## ----------- Network Parameters -----------
+        self.conv_layer_1 = BasicConv(in_dim, self.inter_dim, kernel_size=1, act_type=cfg.neck_act, norm_type=cfg.neck_norm)
+        self.conv_layer_2 = BasicConv(self.inter_dim * 4, self.out_dim, kernel_size=1, act_type=cfg.neck_act, norm_type=cfg.neck_norm)
+        self.pool_layer   = nn.MaxPool2d(kernel_size=cfg.spp_pooling_size, stride=1, padding=cfg.spp_pooling_size // 2)
+
+    def forward(self, x):
+        y = [self.conv_layer_1(x)]
+        y.extend(self.pool_layer(y[-1]) for _ in range(3))
+        
+        return self.conv_layer_2(torch.cat(y, 1))
+    

models/gelan/gelan_pafpn.py

@@ -0,0 +1,103 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from typing import List
+
+from .gelan_basic import RepGElanLayer, ADown
+
+
+# PaFPN-ELAN
+class GElanPaFPN(nn.Module):
+    def __init__(self,
+                 cfg,
+                 in_dims :List = [256, 512, 256],
+                 ) -> None:
+        super(GElanPaFPN, self).__init__()
+        print('==============================')
+        print('FPN: {}'.format("GELAN PaFPN"))
+        # --------------------------- Basic Parameters ---------------------------
+        self.in_dims = in_dims[::-1]
+        self.out_dims = [cfg.fpn_feats_td["p3"][1], cfg.fpn_feats_bu["p4"][1], cfg.fpn_feats_bu["p5"][1]]
+
+        # ---------------- Top dwon ----------------
+        ## P5 -> P4
+        self.top_down_layer_1 = RepGElanLayer(in_dim     = self.in_dims[0] + self.in_dims[1],
+                                              inter_dims = cfg.fpn_feats_td["p4"][0],
+                                              out_dim    = cfg.fpn_feats_td["p4"][1],
+                                              num_blocks = cfg.fpn_depth,
+                                              shortcut   = False,
+                                              act_type   = cfg.fpn_act,
+                                              norm_type  = cfg.fpn_norm,
+                                              depthwise  = cfg.fpn_depthwise,
+                                              )
+        ## P4 -> P3
+        self.top_down_layer_2 = RepGElanLayer(in_dim     = cfg.fpn_feats_td["p4"][1] + self.in_dims[2],
+                                              inter_dims = cfg.fpn_feats_td["p3"][0],
+                                              out_dim    = cfg.fpn_feats_td["p3"][1],
+                                              num_blocks = cfg.fpn_depth,
+                                              shortcut   = False,
+                                              act_type   = cfg.fpn_act,
+                                              norm_type  = cfg.fpn_norm,
+                                              depthwise  = cfg.fpn_depthwise,
+                                              )
+        # ---------------- Bottom up ----------------
+        ## P3 -> P4
+        self.dowmsample_layer_1 = ADown(cfg.fpn_feats_td["p3"][1], cfg.fpn_feats_td["p3"][1],
+                                        act_type=cfg.fpn_act, norm_type=cfg.fpn_norm, depthwise=cfg.fpn_depthwise, use_pooling=cfg.fpn_down_pooling)
+        self.bottom_up_layer_1  = RepGElanLayer(in_dim     = cfg.fpn_feats_td["p3"][1] + cfg.fpn_feats_td["p4"][1],
+                                                inter_dims = cfg.fpn_feats_bu["p4"][0],
+                                                out_dim    = cfg.fpn_feats_bu["p4"][1],
+                                                num_blocks = cfg.fpn_depth,
+                                                shortcut   = False,
+                                                act_type   = cfg.fpn_act,
+                                                norm_type  = cfg.fpn_norm,
+                                                depthwise  = cfg.fpn_depthwise,
+                                                )
+        ## P4 -> P5
+        self.dowmsample_layer_2 = ADown(cfg.fpn_feats_bu["p4"][1], cfg.fpn_feats_bu["p4"][1],
+                                        act_type=cfg.fpn_act, norm_type=cfg.fpn_norm, depthwise=cfg.fpn_depthwise, use_pooling=cfg.fpn_down_pooling)
+        self.bottom_up_layer_2  = RepGElanLayer(in_dim     = cfg.fpn_feats_td["p4"][1] + self.in_dims[0],
+                                                inter_dims = cfg.fpn_feats_bu["p5"][0],
+                                                out_dim    = cfg.fpn_feats_bu["p5"][1],
+                                                num_blocks = cfg.fpn_depth,
+                                                shortcut   = False,
+                                                act_type   = cfg.fpn_act,
+                                                norm_type  = cfg.fpn_norm,
+                                                depthwise  = cfg.fpn_depthwise,
+                                                )
+        
+        self.init_weights()
+        
+    def init_weights(self):
+        """Initialize the parameters."""
+        for m in self.modules():
+            if isinstance(m, torch.nn.Conv2d):
+                # In order to be consistent with the source code,
+                # reset the Conv2d initialization parameters
+                m.reset_parameters()
+
+    def forward(self, features):
+        c3, c4, c5 = features
+
+        # ------------------ Top down FPN ------------------
+        ## P5 -> P4
+        p5_up = F.interpolate(c5, scale_factor=2.0)
+        p4 = self.top_down_layer_1(torch.cat([p5_up, c4], dim=1))
+
+        ## P4 -> P3
+        p4_up = F.interpolate(p4, scale_factor=2.0)
+        p3 = self.top_down_layer_2(torch.cat([p4_up, c3], dim=1))
+
+        # ------------------ Bottom up FPN ------------------
+        ## p3 -> P4
+        p3_ds = self.dowmsample_layer_1(p3)
+        p4 = self.bottom_up_layer_1(torch.cat([p3_ds, p4], dim=1))
+
+        ## P4 -> 5
+        p4_ds = self.dowmsample_layer_2(p4)
+        p5 = self.bottom_up_layer_2(torch.cat([p4_ds, c5], dim=1))
+
+        out_feats = [p3, p4, p5] # [P3, P4, P5]
+
+        return out_feats
+    

models/gelan/gelan_pred.py

@@ -0,0 +1,152 @@
+import math
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+
+# Single-level pred layer
+class SingleLevelPredLayer(nn.Module):
+    def __init__(self,
+                 cls_dim     :int = 256,
+                 reg_dim     :int = 256,
+                 stride      :int = 32,
+                 reg_max     :int = 16,
+                 num_classes :int = 80,
+                 num_coords  :int = 4):
+        super().__init__()
+        # --------- Basic Parameters ----------
+        self.stride = stride
+        self.cls_dim = cls_dim
+        self.reg_dim = reg_dim
+        self.reg_max = reg_max
+        self.num_classes = num_classes
+        self.num_coords = num_coords
+
+        # --------- Network Parameters ----------
+        self.cls_pred = nn.Conv2d(cls_dim, num_classes, kernel_size=1)
+        self.reg_pred = nn.Conv2d(reg_dim, num_coords, kernel_size=1, groups=4)                
+
+        self.init_bias()
+        
+    def init_bias(self):
+        # cls pred bias
+        b = self.cls_pred.bias.view(1, -1)
+        b.data.fill_(math.log(5 / self.num_classes / (640. / self.stride) ** 2))
+        self.cls_pred.bias = torch.nn.Parameter(b.view(-1), requires_grad=True)
+        # reg pred bias
+        b = self.reg_pred.bias.view(-1, )
+        b.data.fill_(1.0)
+        self.reg_pred.bias = torch.nn.Parameter(b.view(-1), requires_grad=True)
+
+    def generate_anchors(self, fmp_size):
+        """
+            fmp_size: (List) [H, W]
+        """
+        # generate grid cells
+        fmp_h, fmp_w = fmp_size
+        anchor_y, anchor_x = torch.meshgrid([torch.arange(fmp_h), torch.arange(fmp_w)])
+        # [H, W, 2] -> [HW, 2]
+        anchors = torch.stack([anchor_x, anchor_y], dim=-1).float().view(-1, 2)
+        anchors += 0.5  # add center offset
+        anchors *= self.stride
+
+        return anchors
+        
+    def forward(self, cls_feat, reg_feat):
+        # pred
+        cls_pred = self.cls_pred(cls_feat)
+        reg_pred = self.reg_pred(reg_feat)
+
+        # generate anchor boxes: [M, 4]
+        B, _, H, W = cls_pred.size()
+        fmp_size = [H, W]
+        anchors = self.generate_anchors(fmp_size)
+        anchors = anchors.to(cls_pred.device)
+        # stride tensor: [M, 1]
+        stride_tensor = torch.ones_like(anchors[..., :1]) * self.stride
+        
+        # [B, C, H, W] -> [B, H, W, C] -> [B, M, C]
+        cls_pred = cls_pred.permute(0, 2, 3, 1).contiguous().view(B, -1, self.num_classes)
+        reg_pred = reg_pred.permute(0, 2, 3, 1).contiguous().view(B, -1, 4*self.reg_max)
+        
+        # output dict
+        outputs = {"pred_cls": cls_pred,            # List(Tensor) [B, M, C]
+                   "pred_reg": reg_pred,            # List(Tensor) [B, M, 4*(reg_max)]
+                   "anchors": anchors,              # List(Tensor) [M, 2]
+                   "strides": self.stride,          # List(Int) = [8, 16, 32]
+                   "stride_tensor": stride_tensor   # List(Tensor) [M, 1]
+                   }
+
+        return outputs
+
+# Multi-level pred layer
+class GElanPredLayer(nn.Module):
+    def __init__(self,
+                 cfg,
+                 cls_dim,
+                 reg_dim,
+                 ):
+        super().__init__()
+        # --------- Basic Parameters ----------
+        self.cfg = cfg
+        self.cls_dim = cls_dim
+        self.reg_dim = reg_dim
+
+        # ----------- Network Parameters -----------
+        ## pred layers
+        self.multi_level_preds = nn.ModuleList(
+            [SingleLevelPredLayer(cls_dim     = cls_dim,
+                                  reg_dim     = reg_dim,
+                                  stride      = cfg.out_stride[level],
+                                  reg_max     = cfg.reg_max,
+                                  num_classes = cfg.num_classes,
+                                  num_coords  = 4 * cfg.reg_max)
+                                  for level in range(cfg.num_levels)
+                                  ])
+        ## proj conv
+        proj_init = torch.arange(cfg.reg_max, dtype=torch.float)
+        self.proj_conv = nn.Conv2d(cfg.reg_max, 1, kernel_size=1, bias=False).requires_grad_(False)
+        self.proj_conv.weight.data[:] = nn.Parameter(proj_init.view([1, cfg.reg_max, 1, 1]), requires_grad=False)
+
+    def forward(self, cls_feats, reg_feats):
+        all_anchors = []
+        all_strides = []
+        all_cls_preds = []
+        all_reg_preds = []
+        all_box_preds = []
+        for level in range(self.cfg.num_levels):
+            # -------------- Single-level prediction --------------
+            outputs = self.multi_level_preds[level](cls_feats[level], reg_feats[level])
+
+            # -------------- Decode bbox --------------
+            B, M = outputs["pred_reg"].shape[:2]
+            # [B, M, 4*(reg_max)] -> [B, M, 4, reg_max]
+            delta_pred = outputs["pred_reg"].reshape([B, M, 4, self.cfg.reg_max])
+            # [B, M, 4, reg_max] -> [B, reg_max, 4, M]
+            delta_pred = delta_pred.permute(0, 3, 2, 1).contiguous()
+            # [B, reg_max, 4, M] -> [B, 1, 4, M]
+            delta_pred = self.proj_conv(F.softmax(delta_pred, dim=1))
+            # [B, 1, 4, M] -> [B, 4, M] -> [B, M, 4]
+            delta_pred = delta_pred.view(B, 4, M).permute(0, 2, 1).contiguous()
+            ## tlbr -> xyxy
+            x1y1_pred = outputs["anchors"][None] - delta_pred[..., :2] * self.cfg.out_stride[level]
+            x2y2_pred = outputs["anchors"][None] + delta_pred[..., 2:] * self.cfg.out_stride[level]
+            box_pred = torch.cat([x1y1_pred, x2y2_pred], dim=-1)
+
+            # collect results
+            all_cls_preds.append(outputs["pred_cls"])
+            all_reg_preds.append(outputs["pred_reg"])
+            all_box_preds.append(box_pred)
+            all_anchors.append(outputs["anchors"])
+            all_strides.append(outputs["stride_tensor"])
+        
+        # output dict
+        outputs = {"pred_cls":      all_cls_preds,         # List(Tensor) [B, M, C]
+                   "pred_reg":      all_reg_preds,         # List(Tensor) [B, M, 4*(reg_max)]
+                   "pred_box":      all_box_preds,         # List(Tensor) [B, M, 4]
+                   "anchors":       all_anchors,           # List(Tensor) [M, 2]
+                   "stride_tensor": all_strides,           # List(Tensor) [M, 1]
+                   "strides":       self.cfg.out_stride,   # List(Int) = [8, 16, 32]
+                   }
+
+        return outputs

models/gelan/loss.py

@@ -0,0 +1,187 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from utils.box_ops import bbox2dist, bbox_iou
+from utils.distributed_utils import get_world_size, is_dist_avail_and_initialized
+
+from .matcher import TaskAlignedAssigner
+
+
+class SetCriterion(object):
+    def __init__(self, cfg):
+        # --------------- Basic parameters ---------------
+        self.cfg = cfg
+        self.reg_max = cfg.reg_max
+        self.num_classes = cfg.num_classes
+        # --------------- Loss config ---------------
+        self.loss_cls_weight = cfg.loss_cls
+        self.loss_box_weight = cfg.loss_box
+        self.loss_dfl_weight = cfg.loss_dfl
+        # --------------- Matcher config ---------------
+        self.matcher = TaskAlignedAssigner(num_classes     = cfg.num_classes,
+                                           topk_candidates = cfg.tal_topk_candidates,
+                                           alpha           = cfg.tal_alpha,
+                                           beta            = cfg.tal_beta
+                                           )
+
+    def loss_classes(self, pred_cls, gt_score):
+        # compute bce loss
+        loss_cls = F.binary_cross_entropy_with_logits(pred_cls, gt_score, reduction='none')
+
+        return loss_cls
+    
+    def loss_bboxes(self, pred_box, gt_box, bbox_weight):
+        # regression loss
+        ious = bbox_iou(pred_box, gt_box, xywh=False, CIoU=True)
+        loss_box = (1.0 - ious.squeeze(-1)) * bbox_weight
+
+        return loss_box
+    
+    def loss_dfl(self, pred_reg, gt_box, anchor, stride, bbox_weight=None):
+        # rescale coords by stride
+        gt_box_s = gt_box / stride
+        anchor_s = anchor / stride
+
+        # compute deltas
+        gt_ltrb_s = bbox2dist(anchor_s, gt_box_s, self.reg_max - 1)
+
+        gt_left = gt_ltrb_s.to(torch.long)
+        gt_right = gt_left + 1
+
+        weight_left = gt_right.to(torch.float) - gt_ltrb_s
+        weight_right = 1 - weight_left
+
+        # loss left
+        loss_left = F.cross_entropy(
+            pred_reg.view(-1, self.reg_max),
+            gt_left.view(-1),
+            reduction='none').view(gt_left.shape) * weight_left
+        # loss right
+        loss_right = F.cross_entropy(
+            pred_reg.view(-1, self.reg_max),
+            gt_right.view(-1),
+            reduction='none').view(gt_left.shape) * weight_right
+
+        loss_dfl = (loss_left + loss_right).mean(-1)
+        
+        if bbox_weight is not None:
+            loss_dfl *= bbox_weight
+
+        return loss_dfl
+
+    def __call__(self, outputs, targets):        
+        """
+            outputs['pred_cls']: List(Tensor) [B, M, C]
+            outputs['pred_reg']: List(Tensor) [B, M, 4*(reg_max+1)]
+            outputs['pred_box']: List(Tensor) [B, M, 4]
+            outputs['anchors']: List(Tensor) [M, 2]
+            outputs['strides']: List(Int) [8, 16, 32] output stride
+            outputs['stride_tensor']: List(Tensor) [M, 1]
+            targets: (List) [dict{'boxes': [...], 
+                                 'labels': [...], 
+                                 'orig_size': ...}, ...]
+        """
+        # preds: [B, M, C]
+        cls_preds = torch.cat(outputs['pred_cls'], dim=1)
+        reg_preds = torch.cat(outputs['pred_reg'], dim=1)
+        box_preds = torch.cat(outputs['pred_box'], dim=1)
+        bs, num_anchors = cls_preds.shape[:2]
+        device = cls_preds.device
+        anchors = torch.cat(outputs['anchors'], dim=0)
+        
+        # --------------- label assignment ---------------
+        gt_score_targets = []
+        gt_bbox_targets = []
+        fg_masks = []
+        for batch_idx in range(bs):
+            tgt_labels = targets[batch_idx]["labels"].to(device)     # [Mp,]
+            tgt_boxs = targets[batch_idx]["boxes"].to(device)        # [Mp, 4]
+
+            if self.cfg.normalize_coords:
+                img_h, img_w = outputs['image_size']
+                tgt_boxs[..., [0, 2]] *= img_w
+                tgt_boxs[..., [1, 3]] *= img_h
+            
+            if self.cfg.box_format == 'xywh':
+                tgt_boxs_x1y1 = tgt_boxs[..., :2] - 0.5 * tgt_boxs[..., 2:]
+                tgt_boxs_x2y2 = tgt_boxs[..., :2] + 0.5 * tgt_boxs[..., 2:]
+                tgt_boxs = torch.cat([tgt_boxs_x1y1, tgt_boxs_x2y2], dim=-1)
+
+            # check target
+            if len(tgt_labels) == 0 or tgt_boxs.max().item() == 0.:
+                # There is no valid gt
+                fg_mask  = cls_preds.new_zeros(1, num_anchors).bool()               #[1, M,]
+                gt_score = cls_preds.new_zeros((1, num_anchors, self.num_classes)) #[1, M, C]
+                gt_box   = cls_preds.new_zeros((1, num_anchors, 4))                  #[1, M, 4]
+            else:
+                tgt_labels = tgt_labels[None, :, None]      # [1, Mp, 1]
+                tgt_boxs = tgt_boxs[None]                   # [1, Mp, 4]
+                (
+                    _,
+                    gt_box,     # [1, M, 4]
+                    gt_score,   # [1, M, C]
+                    fg_mask,    # [1, M,]
+                    _
+                ) = self.matcher(
+                    pd_scores = cls_preds[batch_idx:batch_idx+1].detach().sigmoid(), 
+                    pd_bboxes = box_preds[batch_idx:batch_idx+1].detach(),
+                    anc_points = anchors,
+                    gt_labels = tgt_labels,
+                    gt_bboxes = tgt_boxs
+                    )
+            gt_score_targets.append(gt_score)
+            gt_bbox_targets.append(gt_box)
+            fg_masks.append(fg_mask)
+
+        # List[B, 1, M, C] -> Tensor[B, M, C] -> Tensor[BM, C]
+        fg_masks = torch.cat(fg_masks, 0).view(-1)                                    # [BM,]
+        gt_score_targets = torch.cat(gt_score_targets, 0).view(-1, self.num_classes)  # [BM, C]
+        gt_bbox_targets = torch.cat(gt_bbox_targets, 0).view(-1, 4)                   # [BM, 4]
+        num_fgs = gt_score_targets.sum()
+        
+        # Average loss normalizer across all the GPUs
+        if is_dist_avail_and_initialized():
+            torch.distributed.all_reduce(num_fgs)
+        num_fgs = (num_fgs / get_world_size()).clamp(1.0)
+
+        # ------------------ Classification loss ------------------
+        cls_preds = cls_preds.view(-1, self.num_classes)
+        loss_cls = self.loss_classes(cls_preds, gt_score_targets)
+        loss_cls = loss_cls.sum() / num_fgs
+
+        # ------------------ Regression loss ------------------
+        box_preds_pos = box_preds.view(-1, 4)[fg_masks]
+        box_targets_pos = gt_bbox_targets.view(-1, 4)[fg_masks]
+        bbox_weight = gt_score_targets[fg_masks].sum(-1)
+        loss_box = self.loss_bboxes(box_preds_pos, box_targets_pos, bbox_weight)
+        loss_box = loss_box.sum() / num_fgs
+
+        # ------------------ Distribution focal loss  ------------------
+        ## process anchors
+        anchors = anchors[None].repeat(bs, 1, 1).view(-1, 2)
+        ## process stride tensors
+        strides = torch.cat(outputs['stride_tensor'], dim=0)
+        strides = strides.unsqueeze(0).repeat(bs, 1, 1).view(-1, 1)
+        ## fg preds
+        reg_preds_pos = reg_preds.view(-1, 4*self.reg_max)[fg_masks]
+        anchors_pos = anchors[fg_masks]
+        strides_pos = strides[fg_masks]
+        ## compute dfl
+        loss_dfl = self.loss_dfl(reg_preds_pos, box_targets_pos, anchors_pos, strides_pos, bbox_weight)
+        loss_dfl = loss_dfl.sum() / num_fgs
+
+        # total loss
+        losses = loss_cls * self.loss_cls_weight + loss_box * self.loss_box_weight + loss_dfl * self.loss_dfl_weight
+        loss_dict = dict(
+                loss_cls = loss_cls,
+                loss_box = loss_box,
+                loss_dfl = loss_dfl,
+                losses = losses
+        )
+
+        return loss_dict
+    
+
+if __name__ == "__main__":
+    pass

models/gelan/matcher.py

@@ -0,0 +1,199 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from utils.box_ops import bbox_iou
+
+
+# -------------------------- Task Aligned Assigner --------------------------
+class TaskAlignedAssigner(nn.Module):
+    def __init__(self,
+                 num_classes     = 80,
+                 topk_candidates = 10,
+                 alpha           = 0.5,
+                 beta            = 6.0, 
+                 eps             = 1e-9):
+        super(TaskAlignedAssigner, self).__init__()
+        self.topk_candidates = topk_candidates
+        self.num_classes = num_classes
+        self.bg_idx = num_classes
+        self.alpha = alpha
+        self.beta = beta
+        self.eps = eps
+
+    @torch.no_grad()
+    def forward(self,
+                pd_scores,
+                pd_bboxes,
+                anc_points,
+                gt_labels,
+                gt_bboxes):
+        self.bs = pd_scores.size(0)
+        self.n_max_boxes = gt_bboxes.size(1)
+
+        mask_pos, align_metric, overlaps = self.get_pos_mask(
+            pd_scores, pd_bboxes, gt_labels, gt_bboxes, anc_points)
+
+        target_gt_idx, fg_mask, mask_pos = select_highest_overlaps(
+            mask_pos, overlaps, self.n_max_boxes)
+
+        # Assigned target
+        target_labels, target_bboxes, target_scores = self.get_targets(
+            gt_labels, gt_bboxes, target_gt_idx, fg_mask)
+
+        # normalize
+        align_metric *= mask_pos
+        pos_align_metrics = align_metric.amax(axis=-1, keepdim=True)  # b, max_num_obj
+        pos_overlaps = (overlaps * mask_pos).amax(axis=-1, keepdim=True)  # b, max_num_obj
+        norm_align_metric = (align_metric * pos_overlaps / (pos_align_metrics + self.eps)).amax(-2).unsqueeze(-1)
+        target_scores = target_scores * norm_align_metric
+
+        return target_labels, target_bboxes, target_scores, fg_mask.bool(), target_gt_idx
+
+    def get_pos_mask(self, pd_scores, pd_bboxes, gt_labels, gt_bboxes, anc_points):
+        # get in_gts mask, (b, max_num_obj, h*w)
+        mask_in_gts = select_candidates_in_gts(anc_points, gt_bboxes)
+        # get anchor_align metric, (b, max_num_obj, h*w)
+        align_metric, overlaps = self.get_box_metrics(pd_scores, pd_bboxes, gt_labels, gt_bboxes, mask_in_gts)
+        # get topk_metric mask, (b, max_num_obj, h*w)
+        mask_topk = self.select_topk_candidates(align_metric)
+        # merge all mask to a final mask, (b, max_num_obj, h*w)
+        mask_pos = mask_topk * mask_in_gts
+
+        return mask_pos, align_metric, overlaps
+
+    def get_box_metrics(self, pd_scores, pd_bboxes, gt_labels, gt_bboxes, mask_in_gts):
+        """Compute alignment metric given predicted and ground truth bounding boxes."""
+        na = pd_bboxes.shape[-2]
+        mask_in_gts = mask_in_gts.bool()  # b, max_num_obj, h*w
+        overlaps = torch.zeros([self.bs, self.n_max_boxes, na], dtype=pd_bboxes.dtype, device=pd_bboxes.device)
+        bbox_scores = torch.zeros([self.bs, self.n_max_boxes, na], dtype=pd_scores.dtype, device=pd_scores.device)
+
+        ind = torch.zeros([2, self.bs, self.n_max_boxes], dtype=torch.long)  # 2, b, max_num_obj
+        ind[0] = torch.arange(end=self.bs).view(-1, 1).expand(-1, self.n_max_boxes)  # b, max_num_obj
+        ind[1] = gt_labels.squeeze(-1)  # b, max_num_obj
+        # Get the scores of each grid for each gt cls
+        bbox_scores[mask_in_gts] = pd_scores[ind[0], :, ind[1]][mask_in_gts]  # b, max_num_obj, h*w
+
+        # (b, max_num_obj, 1, 4), (b, 1, h*w, 4)
+        pd_boxes = pd_bboxes.unsqueeze(1).expand(-1, self.n_max_boxes, -1, -1)[mask_in_gts]
+        gt_boxes = gt_bboxes.unsqueeze(2).expand(-1, -1, na, -1)[mask_in_gts]
+        overlaps[mask_in_gts] = bbox_iou(gt_boxes, pd_boxes, xywh=False, CIoU=True).squeeze(-1).clamp_(0)
+
+        align_metric = bbox_scores.pow(self.alpha) * overlaps.pow(self.beta)
+        return align_metric, overlaps
+
+    def select_topk_candidates(self, metrics, largest=True):
+        """
+        Args:
+            metrics: (b, max_num_obj, h*w).
+            topk_mask: (b, max_num_obj, topk) or None
+        """
+        # (b, max_num_obj, topk)
+        topk_metrics, topk_idxs = torch.topk(metrics, self.topk_candidates, dim=-1, largest=largest)
+        topk_mask = (topk_metrics.max(-1, keepdim=True)[0] > self.eps).expand_as(topk_idxs)
+        # (b, max_num_obj, topk)
+        topk_idxs.masked_fill_(~topk_mask, 0)
+
+        # (b, max_num_obj, topk, h*w) -> (b, max_num_obj, h*w)
+        count_tensor = torch.zeros(metrics.shape, dtype=torch.int8, device=topk_idxs.device)
+        ones = torch.ones_like(topk_idxs[:, :, :1], dtype=torch.int8, device=topk_idxs.device)
+        for k in range(self.topk_candidates):
+            # Expand topk_idxs for each value of k and add 1 at the specified positions
+            count_tensor.scatter_add_(-1, topk_idxs[:, :, k:k + 1], ones)
+        # count_tensor.scatter_add_(-1, topk_idxs, torch.ones_like(topk_idxs, dtype=torch.int8, device=topk_idxs.device))
+        # Filter invalid bboxes
+        count_tensor.masked_fill_(count_tensor > 1, 0)
+
+        return count_tensor.to(metrics.dtype)
+
+    def get_targets(self, gt_labels, gt_bboxes, target_gt_idx, fg_mask):
+        # Assigned target labels, (b, 1)
+        batch_ind = torch.arange(end=self.bs, dtype=torch.int64, device=gt_labels.device)[..., None]
+        target_gt_idx = target_gt_idx + batch_ind * self.n_max_boxes  # (b, h*w)
+        target_labels = gt_labels.long().flatten()[target_gt_idx]  # (b, h*w)
+
+        # Assigned target boxes, (b, max_num_obj, 4) -> (b, h*w, 4)
+        target_bboxes = gt_bboxes.view(-1, 4)[target_gt_idx]
+
+        # Assigned target scores
+        target_labels.clamp_(0)
+
+        # 10x faster than F.one_hot()
+        target_scores = torch.zeros((target_labels.shape[0], target_labels.shape[1], self.num_classes),
+                                    dtype=torch.int64,
+                                    device=target_labels.device)  # (b, h*w, 80)
+        target_scores.scatter_(2, target_labels.unsqueeze(-1), 1)
+
+        fg_scores_mask = fg_mask[:, :, None].repeat(1, 1, self.num_classes)  # (b, h*w, 80)
+        target_scores = torch.where(fg_scores_mask > 0, target_scores, 0)
+
+        return target_labels, target_bboxes, target_scores
+    
+
+# -------------------------- Basic Functions --------------------------
+def select_candidates_in_gts(xy_centers, gt_bboxes, eps=1e-9):
+    """select the positive anchors's center in gt
+    Args:
+        xy_centers (Tensor): shape(bs*n_max_boxes, num_total_anchors, 4)
+        gt_bboxes (Tensor): shape(bs, n_max_boxes, 4)
+    Return:
+        (Tensor): shape(bs, n_max_boxes, num_total_anchors)
+    """
+    n_anchors = xy_centers.size(0)
+    bs, n_max_boxes, _ = gt_bboxes.size()
+    _gt_bboxes = gt_bboxes.reshape([-1, 4])
+    xy_centers = xy_centers.unsqueeze(0).repeat(bs * n_max_boxes, 1, 1)
+    gt_bboxes_lt = _gt_bboxes[:, 0:2].unsqueeze(1).repeat(1, n_anchors, 1)
+    gt_bboxes_rb = _gt_bboxes[:, 2:4].unsqueeze(1).repeat(1, n_anchors, 1)
+    b_lt = xy_centers - gt_bboxes_lt
+    b_rb = gt_bboxes_rb - xy_centers
+    bbox_deltas = torch.cat([b_lt, b_rb], dim=-1)
+    bbox_deltas = bbox_deltas.reshape([bs, n_max_boxes, n_anchors, -1])
+    return (bbox_deltas.min(axis=-1)[0] > eps).to(gt_bboxes.dtype)
+
+def select_highest_overlaps(mask_pos, overlaps, n_max_boxes):
+    """if an anchor box is assigned to multiple gts,
+        the one with the highest iou will be selected.
+    Args:
+        mask_pos (Tensor): shape(bs, n_max_boxes, num_total_anchors)
+        overlaps (Tensor): shape(bs, n_max_boxes, num_total_anchors)
+    Return:
+        target_gt_idx (Tensor): shape(bs, num_total_anchors)
+        fg_mask (Tensor): shape(bs, num_total_anchors)
+        mask_pos (Tensor): shape(bs, n_max_boxes, num_total_anchors)
+    """
+    fg_mask = mask_pos.sum(-2)
+    if fg_mask.max() > 1:  # one anchor is assigned to multiple gt_bboxes
+        mask_multi_gts = (fg_mask.unsqueeze(1) > 1).expand(-1, n_max_boxes, -1)  # (b, n_max_boxes, h*w)
+        max_overlaps_idx = overlaps.argmax(1)  # (b, h*w)
+
+        is_max_overlaps = torch.zeros(mask_pos.shape, dtype=mask_pos.dtype, device=mask_pos.device)
+        is_max_overlaps.scatter_(1, max_overlaps_idx.unsqueeze(1), 1)
+
+        mask_pos = torch.where(mask_multi_gts, is_max_overlaps, mask_pos).float()  # (b, n_max_boxes, h*w)
+        fg_mask = mask_pos.sum(-2)
+    # Find each grid serve which gt(index)
+    target_gt_idx = mask_pos.argmax(-2)  # (b, h*w)
+
+    return target_gt_idx, fg_mask, mask_pos
+
+def iou_calculator(box1, box2, eps=1e-9):
+    """Calculate iou for batch
+    Args:
+        box1 (Tensor): shape(bs, n_max_boxes, 1, 4)
+        box2 (Tensor): shape(bs, 1, num_total_anchors, 4)
+    Return:
+        (Tensor): shape(bs, n_max_boxes, num_total_anchors)
+    """
+    box1 = box1.unsqueeze(2)  # [N, M1, 4] -> [N, M1, 1, 4]
+    box2 = box2.unsqueeze(1)  # [N, M2, 4] -> [N, 1, M2, 4]
+    px1y1, px2y2 = box1[:, :, :, 0:2], box1[:, :, :, 2:4]
+    gx1y1, gx2y2 = box2[:, :, :, 0:2], box2[:, :, :, 2:4]
+    x1y1 = torch.maximum(px1y1, gx1y1)
+    x2y2 = torch.minimum(px2y2, gx2y2)
+    overlap = (x2y2 - x1y1).clip(0).prod(-1)
+    area1 = (px2y2 - px1y1).clip(0).prod(-1)
+    area2 = (gx2y2 - gx1y1).clip(0).prod(-1)
+    union = area1 + area2 - overlap + eps
+
+    return overlap / union

models/yolov6/yolov6.py

@@ -43,6 +43,13 @@ class Yolov6(nn.Module):
         ## Pred
         self.pred     = Yolov6DetPredLayer(cfg, self.fpn.out_dims)
 
+        self.switch_deploy()
+
+    def switch_deploy(self,):
+        for m in self.modules():
+            if hasattr(m, "switch_to_deploy"):
+                m.switch_to_deploy()
+
     def post_process(self, cls_preds, box_preds):
         """
         We process predictions at each scale hierarchically