fjfabz / tf-keras-yolov2-tracking

yolo v2 implementation for object detection and object tracking(SORT) with tensorflow and keras

yolov2

yolo v2 is an object detection algorithm using deep learning.

This repository contains learning and testing in keras and tensorflow.

ToDo

• refactor trainning code
• tracking function(SORT and Deep SORT) based object detection of yolov2
• demo gif

Prerequirement

• python3.6
• opencv for python2 involving movie function
• dataset(image and annotation using XML format)
• keras(tensorflow)

Usage

train

If you want to learn with your dataset you need to edit config.json.

Please prepare the annotation in XML format with reference to yolov 2 formula.

python train.py -c config.json

test

python predict.py -c config.json -w path/to/pretrained_weights -i path/to/image -o path/to/output_dir

tracking

todo

Tracking code performed by SORT algorithm using Kalman filter.

DEMO

todo