vehicle-counting-yolov4-deepsort

Vehicle Counting and Velocity Estimation by YOLO v4 and DeepSORT

Requirements

CUDA Toolkit version 10.1 https://developer.nvidia.com/cuda-10.1-download-archive-update2

Install the proper dependencies either via Anaconda or Pip.

Conda

# Tensorflow CPU
conda env create -f conda-cpu.yml
conda activate yolov4-cpu

# Tensorflow GPU
conda env create -f conda-gpu.yml
conda activate yolov4-gpu

Pip

# TensorFlow CPU
pip install -r requirements.txt

# TensorFlow GPU
pip install -r requirements-gpu.txt

Sample Video

Download sample video at here, put the video file into the 'data/video' folder of this repository.

Run the Vehicle Counting Code with YOLOv4 and DeepSORT

# Convert darknet weights to tensorflow model
python save_model.py --model yolov4 

# Run the Vehicle Counting Code with YOLOv4 and DeepSORT on video
python vehicle_count_velocity.py --video ./data/video/MVI_2966.MP4 --output ./outputs/output.mp4 --model yolov4

References:

yolov4-deepsort by theAIGuysCode