mnn_tracking_deepsort

Python code run MNN face detection and tracking using DeepSort

Install and run this project

Step 1: Clone the repository:

git clone https://github.com/anhtuanlinux/mnn_tracking_deepsort.git

Step 2: CD to the main folder

cd mnn_tracking_deepsort/MNN/python

Step 2: Install libraries and all requirements dependencies:

sudo apt install graphviz

pip install -U pip

pip install -U MNN

pip install -U Cython

pip install -r requirements.txt

Step 3: Run MNN face detection + DeepSORT tracking code

bash run_tracking_ultraface_mnn.sh

where --model_path: MNN model file path

  --source: input video file path
  
  --config_deepsort: yaml file that storing deepsort config
       
  --deep_sort_model: reid checkpoint model file path
  
  --save-vid (True) for saving output video
  
  --save-txt (True) for saving output text file (MOT output format)

Result:

Output video is saved at mnn_tracking_deepsort/MNN/python/runs/track/face_detect_osnet_ibn_x1_0_MSMT17/short_video_test_02.mp4

Output MOT format text file is saved at mnn_tracking_deepsort/MNN/python/runs/track/face_detect_osnet_ibn_x1_0_MSMT17/tracks/short_video_test_02.txt