WiraDKP / cow_detection_opencv

Cow Detection using OpenCV

This project is created as celebration of Eid al-Adha 2020. The object detection is performed with Pretrained SSD on COCO dataset. The SSD uses mobilenet v2 as backbone. The video is royalty free and you can found it here

Step 1: Clone this repository

git clone https://github.com/WiraDKP/cow_detection_opencv.git

Step 2: Setup conda environment

Please download miniconda if you haven't. Then use this command to install the environment

cd /path/to/cow_detection_opencv
conda env create -f environment.yml

Step 3: Upload your video and setup config

You may use your own video. Upload it to the data folder. Then change FILE_PATH in config.py.
You may also change other configuration as you need.

FILE_PATH = "data/cows.mp4"
MODEL_PATH = "model/mobilenet_v2_ssd_coco_frozen_graph.pb"
CONFIG_PATH = "model/mobilenet_v2_ssd_coco_config.pbtxt"
LABEL_PATH = "model/coco_class_labels.txt"

WINDOW_NAME = "detection"

MIN_CONF = 0.4
MAX_IOU = 0.5

Step 4: Activate environment and run the script

conda activate cow_detection
python main.py