Implementation of YOLO tiny (version 1) with tensorflow. This project only supports test with image or video. Please use darknet for training.
- model: tiny yolo checkpoint
- filetype: 'image' or 'video'
- testdir: dictionary for test files
- prob_threshold: probability threshold for bounding boxes, default 0.1
- NMS_threshold: IOU threshold for NMS, default 0.5
- show: show the output image/video
- output_txt: write detection results into text
- output_img: write detection results into image/video (only .avi is supported)
e.g.
python run_test.py --model='weigts/YOLO_tiny.ckpt' --filetype='image' --testdir='test_imgs' --output_txt --output_img --show
- Python 3.6
- Tensorflow 1.6
- opencv-python 3.4
More details in the future...