train YOLOvx-M on COCO

README.md

@@ -151,6 +151,15 @@ python train.py --cuda -d coco --root path/to/COCO -m yolov1 -bs 16 --max_epoch
 
 *While YOLOv7 incorporates several technical details, such as anchor box, SimOTA, AuxiliaryHead, and RepConv, I found it too challenging to fully reproduce. Instead, I created a simpler version of YOLOv7 using an anchor-free structure and SimOTA. As a result, my reproduction had poor performance due to the absence of the other technical details. However, since it was only intended as a tutorial, I am not too concerned about this gap.*
 
+* My YOLO:
+
+| Model    | Scale | Epoch | AP<sup>test<br>0.5:0.95 | AP<sup>val<br>0.5:0.95 | AP<sup>val<br>0.5 | FLOPs<br><sup>(G) | Params<br><sup>(M) | Weight |
+|----------|-------|-------|-------------------------|------------------------|-------------------|-------------------|--------------------|--------|
+| YOLOvx-N |  640  |  300  |                         |                        |                   |                   |                    |  |
+| YOLOvx-S |  640  |  300  |                         |                        |                   |                   |                    |  |
+| YOLOvx-M |  640  |  300  |                         |                        |                   |                   |                    |  |
+| YOLOvx-L |  640  |  300  |                         |          50.0          |        68.4       |      176.6        |        47.6        | [ckpt](https://github.com/yjh0410/PyTorch_YOLO_Tutorial/releases/download/yolo_tutorial_ckpt/yolovx_l_coco.pth) |
+
 * Redesigned RT-DETR:
 
 | Model     | Scale | Epoch | AP<sup>val<br>0.5:0.95 | AP<sup>val<br>0.5 | FLOPs<br><sup>(G) | Params<br><sup>(M) | Weight |

config/data_config/transform_config.py

@@ -205,7 +205,7 @@ yolox_nano_trans_config = {
     'hsv_v': 0.4,
     # Mosaic & Mixup
     'mosaic_prob': 1.0,
-    'mixup_prob': 0.5,
+    'mixup_prob': 0.0,
     'mosaic_type': 'yolov5_mosaic',
     'mixup_type': 'yolox_mixup',
     'mixup_scale': [0.5, 1.5]   # "mixup_scale" is not used for YOLOv5MixUp
@@ -240,112 +240,3 @@ ssd_trans_config = {
     'mixup_type': 'yolov5_mixup',
     'mixup_scale': [0.5, 1.5]
 }
-
-# ----------------------- RTMDet-Style Transform -----------------------
-rtmdet_huge_trans_config = {
-    'aug_type': 'yolov5',
-    # Basic Augment
-    'degrees': 0.0,
-    'translate': 0.2,
-    'scale': 0.9,
-    'shear': 0.0,
-    'perspective': 0.0,
-    'hsv_h': 0.015,
-    'hsv_s': 0.7,
-    'hsv_v': 0.4,
-    # Mosaic & Mixup
-    'mosaic_prob': 1.0,
-    'mixup_prob': 1.0,
-    'mosaic_type': 'yolov5_mosaic',
-    'mixup_type': 'yolov5_mixup',
-}
-
-rtmdet_large_trans_config = {
-    'aug_type': 'yolov5',
-    # Basic Augment
-    'degrees': 0.0,
-    'translate': 0.2,
-    'scale': 0.9,
-    'shear': 0.0,
-    'perspective': 0.0,
-    'hsv_h': 0.015,
-    'hsv_s': 0.7,
-    'hsv_v': 0.4,
-    # Mosaic & Mixup
-    'mosaic_prob': 1.0,
-    'mixup_prob': 1.0,
-    'mosaic_type': 'yolov5_mosaic',
-    'mixup_type': 'yolov5_mixup',
-}
-
-rtmdet_medium_trans_config = {
-    'aug_type': 'yolov5',
-    # Basic Augment
-    'degrees': 0.0,
-    'translate': 0.2,
-    'scale': 0.9,
-    'shear': 0.0,
-    'perspective': 0.0,
-    'hsv_h': 0.015,
-    'hsv_s': 0.7,
-    'hsv_v': 0.4,
-    # Mosaic & Mixup
-    'mosaic_prob': 1.0,
-    'mixup_prob': 1.0,
-    'mosaic_type': 'yolov5_mosaic',
-    'mixup_type': 'yolov5_mixup',
-}
-
-rtmdet_small_trans_config = {
-    'aug_type': 'yolov5',
-    # Basic Augment
-    'degrees': 0.0,
-    'translate': 0.2,
-    'scale': 0.9,
-    'shear': 0.0,
-    'perspective': 0.0,
-    'hsv_h': 0.015,
-    'hsv_s': 0.7,
-    'hsv_v': 0.4,
-    # Mosaic & Mixup
-    'mosaic_prob': 1.0,
-    'mixup_prob': 1.0,
-    'mosaic_type': 'yolov5_mosaic',
-    'mixup_type': 'yolov5_mixup',
-}
-
-rtmdet_nano_trans_config = {
-    'aug_type': 'yolov5',
-    # Basic Augment
-    'degrees': 0.0,
-    'translate': 0.1,
-    'scale': 0.5,
-    'shear': 0.0,
-    'perspective': 0.0,
-    'hsv_h': 0.015,
-    'hsv_s': 0.7,
-    'hsv_v': 0.4,
-    # Mosaic & Mixup
-    'mosaic_prob': 1.0,
-    'mixup_prob': 0.5,
-    'mosaic_type': 'yolov5_mosaic',
-    'mixup_type': 'yolov5_mixup',
-}
-
-rtmdet_pico_trans_config = {
-    'aug_type': 'yolov5',
-    # Basic Augment
-    'degrees': 0.0,
-    'translate': 0.1,
-    'scale': 0.5,
-    'shear': 0.0,
-    'perspective': 0.0,
-    'hsv_h': 0.015,
-    'hsv_s': 0.7,
-    'hsv_v': 0.4,
-    # Mosaic & Mixup
-    'mosaic_prob': 0.5,
-    'mixup_prob': 0.0,
-    'mosaic_type': 'yolov5_mosaic',
-    'mixup_type': 'yolov5_mixup',
-}

config/model_config/yolovx_config.py

@@ -41,7 +41,7 @@ yolovx_cfg = {
         # ---------------- Train config ----------------
         ## Input
         'multi_scale': [0.5, 1.25],   # 320 -> 800
-        'trans_type': 'rtmdet_pico',
+        'trans_type': 'yolox_pico',
         # ---------------- Assignment config ----------------
         ## Matcher
         'matcher': {'center_sampling_radius': 2.5,
@@ -95,7 +95,7 @@ yolovx_cfg = {
         # ---------------- Train config ----------------
         ## Input
         'multi_scale': [0.5, 1.25],   # 320 -> 800
-        'trans_type': 'rtmdet_nano',
+        'trans_type': 'yolox_nano',
         # ---------------- Assignment config ----------------
         ## Matcher
         'matcher': {'center_sampling_radius': 2.5,
@@ -149,7 +149,7 @@ yolovx_cfg = {
         # ---------------- Train config ----------------
         ## Input
         'multi_scale': [0.5, 1.25],   # 320 -> 800
-        'trans_type': 'rtmdet_nano',
+        'trans_type': 'yolox_nano',
         # ---------------- Assignment config ----------------
         ## Matcher
         'matcher': {'center_sampling_radius': 2.5,
@@ -203,7 +203,7 @@ yolovx_cfg = {
         # ---------------- Train config ----------------
         ## Input
         'multi_scale': [0.5, 1.25],   # 320 -> 800
-        'trans_type': 'rtmdet_small',
+        'trans_type': 'yolox_small',
         # ---------------- Assignment config ----------------
         ## Matcher
         'matcher': {'center_sampling_radius': 2.5,
@@ -257,7 +257,7 @@ yolovx_cfg = {
         # ---------------- Train config ----------------
         ## Input
         'multi_scale': [0.5, 1.25],   # 320 -> 800
-        'trans_type': 'rtmdet_medium',
+        'trans_type': 'yolox_medium',
         # ---------------- Assignment config ----------------
         ## Matcher
         'matcher': {'center_sampling_radius': 2.5,
@@ -311,7 +311,7 @@ yolovx_cfg = {
         # ---------------- Train config ----------------
         ## Input
         'multi_scale': [0.5, 1.25],   # 320 -> 800
-        'trans_type': 'rtmdet_large',
+        'trans_type': 'yolox_large',
         # ---------------- Assignment config ----------------
         ## Matcher
         'matcher': {'center_sampling_radius': 2.5,
@@ -365,7 +365,7 @@ yolovx_cfg = {
         # ---------------- Train config ----------------
         ## Input
         'multi_scale': [0.5, 1.25],   # 320 -> 800
-        'trans_type': 'rtmdet_huge',
+        'trans_type': 'yolox_huge',
         # ---------------- Assignment config ----------------
         ## Matcher
         'matcher': {'center_sampling_radius': 2.5,

train.py

@@ -59,7 +59,7 @@ def parse_args():
     # Model
     parser.add_argument('-m', '--model', default='yolov1', type=str,
                         help='build yolo')
-    parser.add_argument('-ct', '--conf_thresh', default=0.005, type=float,
+    parser.add_argument('-ct', '--conf_thresh', default=0.001, type=float,
                         help='confidence threshold')
     parser.add_argument('-nt', '--nms_thresh', default=0.6, type=float,
                         help='NMS threshold')

train_ddp.sh

@@ -5,7 +5,7 @@ python -m torch.distributed.run --nproc_per_node=8 train.py \
                                                     -dist \
                                                     -d coco \
                                                     --root /data/datasets/ \
-                                                    -m yolovx_l \
+                                                    -m yolovx_m \
                                                     -bs 128 \
                                                     -size 640 \
                                                     --wp_epoch 3 \