train YOLOv5 on VOC

README.md

@@ -72,7 +72,7 @@ python train.py --cuda -d voc --root path/to/VOCdevkit -v yolov1 -bs 16 --max_ep
 | YOLOv4 | CSPDarkNet-53 |  640  |  √   |  150  | 83.6 |                          |   162.7           |   61.5             | [ckpt](https://github.com/yjh0410/PyTorch_YOLO_Tutorial/releases/download/yolo_tutorial_ckpt/yolov4_voc.pth) |
 | YOLOv5 | CSPDarkNet-L  |  640  |  √   |  150  |      |                          |                   |                    |  |
 | YOLOX  | CSPDarkNet-L  |  640  |  √   |  150  | 84.6 |                          |   155.4           |   54.2             | [ckpt](https://github.com/yjh0410/PyTorch_YOLO_Tutorial/releases/download/yolo_tutorial_ckpt/yolox_voc.pth) |
-| YOLOv7 | ELANNet       |  640  |  √   |  150  |      |                          |   144.6           |   44.0             |  |
+| YOLOv7 | ELANNet       |  640  |  √   |  150  | 85.5 |                          |   144.6           |   44.0             |  |
 
 *All models are trained with ImageNet pretrained weight (IP). All FLOPs are measured with a 640x640 image size on VOC2007 test. The FPS is measured with batch size 1 on 3090 GPU from the model inference to the NMS operation.*
 

config/yolov5_config.py

@@ -6,7 +6,7 @@ yolov5_cfg = {
     'multi_scale': [0.5, 1.0],
     # model
     'backbone': 'cspdarknet',
-    'pretrained': False,
+    'pretrained': True,
     'bk_act': 'silu',
     'bk_norm': 'BN',
     'bk_dpw': False,

train.sh

@@ -3,7 +3,7 @@ python train.py \
         --cuda \
         -d voc \
         --root /mnt/share/ssd2/dataset/ \
-        -m yolov7 \
+        -m yolov5 \
         -bs 16 \
         -size 640 \
         --wp_epoch 1 \