Re-implementation PyTorch based YOLOv3Tiny
I tried to reproduce Darknet based YOLOv3Tiny using PyTorch.
I used the ImageNet pretrained backbone network.
Refer to this project to know training rules for training the pretrained network.
| Training Set | Test Set | mAP(mean Average Precision) | Weights |
|---|---|---|---|
| VOC 07+12 train/val | VOC 07 test | 58.86 | download |
Use this script to prepare VOC dataset.
If you finished to download and convert labeling format, put the dataset folder into project folder.
├─backbone_weights
├─figures
├─test_example
├─VOCdataset
│ ├─test
│ └─train
| ...
| train.py
| test.py
python train.py
python test.py --weights your_weight.pth