This is the official implementation of the paper "Cascade-SORT: A Robust Fruit Counting Approach Using Multiple Features Cascade Matching".
At present, This code have beeen verified on MacOS 10.15.6. More functions will be added in future versions, to be continued...
- python 3.8
- numpy 1.18.5
- scipy 1.5.0
- opencv-python 4.4.0.44
- opencv-contrib-python 4.4.0.44
- scikit learn 0.23.1
-
Check all dependencies installed
-
Clone this repository
git clone git@github.com:ZQPei/deep_sort_pytorch.git
- Download the YOLO Weights from the followed links:
Google Drive:
https://drive.google.com/file/d/1lNvWKdFl36FrY-Cj2vEZrx-H8okXkcbT/view?usp=sharing
Baidu:
https://pan.baidu.com/s/1JA5lVb_BkQGbWy_u9bwdug
Extract code: 5efn
- Set configuration, revise "yaml/apple.yaml", if the code is run on the custom videos and models
YOLO:
CFG: "cfg/apple.cfg"
WEIGHT: "checkpoints/apple_best.weights"
CLASS_NAMES: "cfg/apple.names"
SCORE_THRESH: 0.5
NMS_THRESH: 0.5
TRACK:
MODE: "cascade"
VIDEO_DIR: "video/apple.mp4"
SAVE_DIR: "results/apple.mp4"- Run demo
python main.py