tensorflow-yolo3

Detection

1、If use the pretrain model, download YOLOV3 weights from YOLO website.
2、Modify yolo3weightspath in the config.py
3、Run detect.py

wget https://pjreddie.com/media/files/yolov3.weights  
python detect.py --image_file ./test.jpg  

Training

convert train and val data to tfrecord

1、Download the COCO2017 dataset from COCO_website
2、Modify the train and val data path in the config.py
3、If you want to use original pretrained weights for YOLOv3, download from darknet53 weights
4、rename it as darknet53.weights, and modify the darknet53weightspath in the config.py

wget https://pjreddie.com/media/files/darknet53.conv.74`  

4、Modify the data augmentation parameters and train parameters
5、Run yolo_train.py

Evaluation

1、Modify the pretrainyolo3 and model_dir in config.py
2、Run detect.py

python detect.py --image_file ./test.jpg

Train Image show on Tensorboard

Notice

If you want to modify the Gpu index, please modify gpu_index in config.py

Credit

@article{yolov3,
    title={YOLOv3: An Incremental Improvement},
    author={Redmon, Joseph and Farhadi, Ali},
    journal = {arXiv},
    year={2018}
}

Reference

• keras-yolo3