xxx975 / vehicle_tracking

OpenCV 3 & Keras implementation of vehicle tracking with video data.

Vehicle_Tracking

OpenCV 3 & Keras implementation of vehicles tracking with video data.

Algorithm

1. Object Detection: MOG2
2. Object Tracking: KCF
3. Object Classification: CNN

Requirement

• Python 3.6
• OpenCV 3.2 + contrib
• Tensorflow-gpu 1.0
• Keras 1.2

Data

We train our CNN model with MIT's vehicle and pedestrian data, click here to download the original data and the processed data.

Video data is saved in the video folder.

CNN Model

The CNN model we use is as follows：

Result

Run the following command to execute the program.

python track.py --file "car.flv"