Given an input video, detects pedestrians using either HOG (Histogram of Oriented Gradients) or R-CNN's. For performance reasons, only searches regions where movement occurs, using background subtraction.
Tracking is done using Kalman Filter. Report format is MOT compatible. Link: https://motchallenge.net/instructions/
Uses models from: https://github.com/opencv/opencv/wiki/TensorFlow-Object-Detection-API
Requires mobilenet folder containing:
Requires: (Tested on Python 3.7 with Anaconda)
- OpenCV 4.2
- Numpy
- Skimage 0.16.2
- Tensorflow 2.0.0