sapjunior / objdetection-pytorch

YOLOv3 Implementation in PyTorch

YOLOv3 Implementation in PyTorch (WIP)

Required Package

• Python3
• PyTorch >= 0.4.1
• OpenCV
• Imgaug (https://github.com/aleju/imgaug)
• Pillow-SIMD (https://github.com/uploadcare/pillow-simd)

Pretrained Weights & Raccoon Dataset Weights

https://drive.google.com/drive/folders/1Fh1mTowJR1KuVhCTRAnEL28UpKUyD7lx?usp=sharing (1)

How to use?

Training (Network configuration is defined in config-raccoon.py)

python3 train.py config-raccoon.py

Predict (Download racccoon/best.pth from above link and put in trainingLogs/raccoonDataset/best.pth)

python3 detect.py config-raccoon.py datasets/raccoon/raccoon-10.jpg

Available Backbones

• Darknet53
• Darknet21 (WIP)
• NASNetAMobile

TODO:

• Training visualization using TensorboardX
• MAP evaluation
• User friendly documents
• More backbone networks
• CPU only Inference