Object Tracking and Deep Learning

This repo contains the code and results used to evaluate the performance of the visual object tracking algorithm.

Instructions:

1. Download the MDNet source code available at: here.
2. Unzip the file in a directory
3. download the performance_eval folder from this repo and extract it inside the root of the MDNet directory.
4. Download the OTB dataset and save it in the dataset directory.

Usage:

First, run the pre-trained MDNet tracker using

cd tracking
python run_tracker.py -s DragonBaby [-d (display fig)] [-f (save fig)]

To create a video from the image sequences, go inside the performance_eval directory and run the command:

python generateVideo.py -d [Sequence Name] -c [value for the size of the frame counter] -f [frame per second]

To generate the Average overlap Score (AOS) and the success plot run

python performance_eval.py

Note: to generate the report create a text file containing the name of all sequences and save it as seq_list.txt

@InProceedings{nam2016mdnet,
author = {Nam, Hyeonseob and Han, Bohyung},
title = {Learning Multi-Domain Convolutional Neural Networks for Visual Tracking},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2016}
}