Dataset and Code for the paper "DepthTrack: Unveiling the Power of RGBD Tracking" (ICCV2021), and "Depth-only Object Tracking" (BMVC2021)

Last update: Dec 15, 2022

Overview

DeT and DOT

Code and datasets for

"DepthTrack: Unveiling the Power of RGBD Tracking" (ICCV2021)
"Depth-only Object Tracking" (BMVC2021)

@InProceedings{yan2021det,
    author    = {Yan, Song and Yang, Jinyu and Kapyla, Jani and Zheng, Feng and Leonardis, Ales and Kamarainen, Joni-Kristian},
    title     = {DepthTrack: Unveiling the Power of RGBD Tracking},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
    month     = {October},
    year      = {2021},
    pages     = {10725-10733}
}

@InProceedings{yan2021dot,
  title       = {Depth-only Object Tracking},
  author      = {Yan, Song and Yang, Jinyu and Leonardis, Ales and Kamarainen, Joni-Kristian},
  booktitle   = {Procedings of the British Machine Vision Conference (BMVC)},
  year        = {2021},
  organization= {British Machine Vision Association}
}

The settings are same as that of Pytracking, please read the document of Pytracking for details.

Generated Depth

We highly recommend to generate high quality depth data from the existing RGB tracking benchmarks, such as LaSOT, Got10K, TrackingNet, and COCO.

We show the examples of generated depth here. The first row is the results from HighResDepth for LaSOT RGB images, the second and the third are from DenseDepth for Got10K and COCO RGB images, the forth row is for the failure cases in which the targets are too close to the background or floor. The last row is from DenseDepth for CDTB RGB images.

In our paper, we used the DenseDepth monocular depth estimation method. We calculate the Ordinal Error (ORD) on the generated depth for CDTB and our DepthTrack test set, and the mean ORD is about 0.386, which is sufficient for training D or RGBD trackers and we have tested it in our works.

And we also tried the recently HighResDepth from CVPR2021, which also performs very well.

@article{alhashim2018high,
  title={High quality monocular depth estimation via transfer learning},
  author={Alhashim, Ibraheem and Wonka, Peter},
  journal={arXiv preprint arXiv:1812.11941},
  year={2018}
}

@inproceedings{miangoleh2021boosting,
  title={Boosting Monocular Depth Estimation Models to High-Resolution via Content-Adaptive Multi-Resolution Merging},
  author={Miangoleh, S Mahdi H and Dille, Sebastian and Mai, Long and Paris, Sylvain and Aksoy, Yagiz},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={9685--9694},
  year={2021}
}

We will public the generated depth maps one by one.

Generated Depth maps for LaSOT

We manually remove bad sequences, and here are totally 646 sequences (some zip files may be broken, will be updated soon) used the DenseDepth method. Original DenseDepth outputs are in range [0, 1.0], we multiply 2^16. Please check LaSOT for RGB images and groundtruth.

part01, part02, part03, part04, part05, part06, part07, part08, part09, part10

The generated depth maps by using HighResDepth will be uploaded soon.

If you find some excellent methods to generate high quality depth images, please share it.

Architecture

Actually the network architecture is very simple, just adding one ResNet50 feature extractor for Depth input and then merging the RGB and Depth feature maps. Below figures are

the feature maps for RGB, D inputs and the merged RGBD ones,
the network for RGBD DiMP50, and
RGBD ATOM.

Download

Download the training dataset(70 sequences) of VOT2021RGBD Challenge from Zenodo (DepthTrack RGBD Tracking Benchmark) and edit the path in local.py More data will be uploaded soon, we hope to bring a large scale RGBD training dataset.

http://doi.org/10.5281/zenodo.4716441

Download the checkpoints for DeT trackers (in install.sh)

gdown https://drive.google.com/uc\?id\=1djSx6YIRmuy3WFjt9k9ZfI8q343I7Y75 -O pytracking/networks/DeT_DiMP50_Max.pth
gdown https://drive.google.com/uc\?id\=1JW3NnmFhX3ZnEaS3naUA05UaxFz6DLFW -O pytracking/networks/DeT_DiMP50_Mean.pth
gdown https://drive.google.com/uc\?id\=1wcGJc1Xq_7d-y-1nWh6M7RaBC1AixRTu -O pytracking/networks/DeT_DiMP50_MC.pth
gdown https://drive.google.com/uc\?id\=17IIroLZ0M_ZVuxkGN6pVy4brTpicMrn8 -O pytracking/networks/DeT_DiMP50_DO.pth
gdown https://drive.google.com/uc\?id\=17aaOiQW-zRCCqPePLQ9u1s466qCtk7Lh -O pytracking/networks/DeT_ATOM_Max.pth
gdown https://drive.google.com/uc\?id\=15LqCjNelRx-pOXAwVd1xwiQsirmiSLmK -O pytracking/networks/DeT_ATOM_Mean.pth
gdown https://drive.google.com/uc\?id\=14wyUaG-pOUu4Y2MPzZZ6_vvtCuxjfYPg -O pytracking/networks/DeT_ATOM_MC.pth

Install

bash install.sh path-to-anaconda DeT

Train

Using the default DiMP50 or ATOM pretrained checkpoints can reduce the training time.

For example, move the default dimp50.pth into the checkpoints folder and rename as DiMPNet_Det_EP0050.pth.tar

python run_training.py bbreg DeT_ATOM_Max
python run_training.py bbreg DeT_ATOM_Mean
python run_training.py bbreg DeT_ATOM_MC

python run_training.py dimp DeT_DiMP50_Max
python run_training.py dimp DeT_DiMP50_Mean
python run_training.py dimp DeT_DiMP50_MC

Test

python run_tracker.py atom DeT_ATOM_Max --dataset_name depthtrack --input_dtype rgbcolormap
python run_tracker.py atom DeT_ATOM_Mean --dataset_name depthtrack --input_dtype rgbcolormap
python run_tracker.py atom DeT_ATOM_MC --dataset_name depthtrack --input_dtype rgbcolormap

python run_tracker.py dimp DeT_DiMP50_Max --dataset_name depthtrack --input_dtype rgbcolormap
python run_tracker.py dimp DeT_DiMP50_Mean --dataset_name depthtrack --input_dtype rgbcolormap
python run_tracker.py dimp DeT_DiMP50_MC --dataset_name depthtrack --input_dtype rgbcolormap


python run_tracker.py dimp dimp50 --dataset_name depthtrack --input_dtype color
python run_tracker.py atom default --dataset_name depthtrack --input_dtype color

Dataset and Code for the paper "DepthTrack: Unveiling the Power of RGBD Tracking" (ICCV2021), and "Depth-only Object Tracking" (BMVC2021)

Related tags

Overview

DeT and DOT

Generated Depth

Generated Depth maps for LaSOT

Architecture

Download

Install

Train

Test

Owner

Yan Song

本步态识别系统主要基于GaitSet模型进行实现

This repository builds a basic vision transformer from scratch so that one beginner can understand the theory of vision transformer.

DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervision

YOLTv5 rapidly detects objects in arbitrarily large aerial or satellite images that far exceed the ~600×600 pixel size typically ingested by deep learning object detection frameworks

minimizer-space de Bruijn graphs (mdBG) for whole genome assembly

Computationally Efficient Optimization of Plackett-Luce Ranking Models for Relevance and Fairness

Official PyTorch Implementation of Hypercorrelation Squeeze for Few-Shot Segmentation, arXiv 2021

Simple Baselines for Human Pose Estimation and Tracking

Trans-Encoder: Unsupervised sentence-pair modelling through self- and mutual-distillations

[TPDS'21] COSCO: Container Orchestration using Co-Simulation and Gradient Based Optimization for Fog Computing Environments

A gesture recognition system powered by OpenPose, k-nearest neighbours, and local outlier factor.

A simple python program that can be used to implement user authentication tokens into your program...

This is the unofficial code of Deep Dual-resolution Networks for Real-time and Accurate Semantic Segmentation of Road Scenes. which achieve state-of-the-art trade-off between accuracy and speed on cityscapes and camvid, without using inference acceleration and extra data

P-Tuning v2: Prompt Tuning Can Be Comparable to Finetuning Universally Across Scales and Tasks

Code for the paper "Curriculum Dropout", ICCV 2017

Realistic lighting in ursina!

Code Release for Learning to Adapt to Evolving Domains

QMagFace: Simple and Accurate Quality-Aware Face Recognition

Pytorch implementation for "Implicit Feature Alignment: Learn to Convert Text Recognizer to Text Spotter".

On Evaluation Metrics for Graph Generative Models