DCFNet

DCFNET: DISCRIMINANT CORRELATION FILTERS NETWORK FOR VISUAL TRACKING_(JCST)

[️‍🔥News️‍🔥] DCFNet is accepted in JCST. If you find DCFNet useful in your research, please consider citing:

@Article{JCST-2309-13788,
title = {DCFNet: Discriminant Correlation Filters Network for Visual Tracking},
journal = {Journal of Computer Science and Technology},
year = {2023},
issn = {1000-9000(Print) /1860-4749(Online)},
doi = {10.1007/s11390-023-3788-3},	
author = {Wei-Ming Hu and Qiang Wang and Jin Gao and Bing Li and Stephen Maybank}
}

Introduction

Discriminant Correlation Filters (DCF) based methods now become a kind of dominant approach to online object tracking. The features used in these methods, however, are either based on hand-crafted features like HoGs, or convolutional features trained independently from other tasks like image classification. In this work, we present an end-to-end lightweight network architecture, namely DCFNet, to learn the convolutional features and perform the correlation tracking process simultaneously.

Contents

1. Requirements
2. Tracking
3. Training
4. Results
5. Citation

Requirements

git clone --depth=1 https://github.com/foolwood/DCFNet.git

Requirements for MatConvNet 1.0-beta24 (see: MatConvNet)

1. Downloading MatConvNet

cd <DCFNet>
git clone https://github.com/vlfeat/matconvnet.git

2. Compiling MatConvNet

Run the following command from the MATLAB command window:

cd matconvnet
run matlab/vl_compilenn

[Optional]

If you want to reproduce the speed in our paper, please follow the website to compile the GPU version.

Tracking

The file demo/demoDCFNet.m is used to test our algorithm.

To reproduce the performance on OTB , you can simple copy DCFNet/ into OTB toolkit.

[Note] Configure MatConvNet path in tracking_env.m

Training

1.Download the training data. (VID)

2.Data Preprocessing in MATLAB.

cd training/dataPreprocessing
data_preprocessing();
analyze_data();

3.Train a DCFNet model.

train_DCFNet();

Results

DCFNet obtains a significant improvements by

• Good Training dataset. (TC128+UAV123+NUS_PRO -> VID)
• Good learning policy. (constant 1e-5 -> logspace(-2,-5,50))
• Large padding size. (1.5 -> 2.0)

The OPE/TRE/SRE results on OTB BaiduYun or GoogleDrive.

Citing DCFNet

If you find DCFNet useful in your research, please consider citing:

@article{wang17dcfnet,
    Author = {Qiang Wang, Jin Gao, Junliang Xing, Mengdan Zhang, Weiming Hu},
    Title = {DCFNet: Discriminant Correlation Filters Network for Visual Tracking},
    Journal = {arXiv preprint arXiv:1704.04057},
    Year = {2017}
}