/YOLO_tiny

Primary LanguagePython

YOLO tiny

Implementation of YOLO tiny (version 1) with tensorflow. This project only supports test with image or video. Please use darknet for training.

Usage

  • 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

Results

dog

Requirements

  • Python 3.6
  • Tensorflow 1.6
  • opencv-python 3.4

References

More details in the future...