Hiring research interns for visual tracking, segmentation and neural architecture search projects: houwen.peng@microsoft.com
💥 Code of OceanPlus has been released!
💥 We achieves the runner-ups for both VOT2020ST (short-term) and RT(real-time). The variants of Ocean take 2nd/3rd/5th places of VOT2020RT. The SiamDW-T submitted to VOT2019 achieves 1st of VOT2020RGBT (submitted by VOT committee).
💥 Our paper Ocean has been accepted by ECCV2020.
💥 The initial version is released, including Ocean(ECCV2020) and SiamDW(CVPR2019).
💥 We provide a TensorRT implementation, running at 1.5~2.5 times faster than pytorch version (e.g. 149fps/68fps for video twinnings
, see details).
Note: We focus on providing an easy-to-follow code based on Pytorch and TensorRT for research on video object tracking and segmentation task. The code will be continuously optimized. You may pull requests to help us build this repo.
🔥 Welcome to subscribe our YouTube Channel.
🔥 Comparision: We summarize the performances of 97 trackers (published in CVPR/ICCV/ECCV/AAAI/NIPS) on 15 tracking benchmarks (OTB13/15, VOT16-20, LASOT, GOT10K, TrackingNet, UAV123, NFS, TC128, VOT2018LT, OxUvA). The repo. is designed to easily compare different trackers, especially when writing papers (performance table/figures). We will continuously update that repo., and we welcome your PR.
[Paper] [Raw Results] [Training and Testing Tutorial] [Demo]
Official implementation of the OceanPlus tracker. It proposes an attention retrieval network (ARN) to perform soft spatial constraints on backbone features. Concretely, we first build a look-up-table (LUT) with the ground-truth mask in the starting frame, and then retrieve the LUT to obtain a target-aware attention map for suppressing the negative influence of pixel-wise background clutter. Furthermore, we introduce a multi-resolution multi-stage segmentation network (MMS) to ulteriorly weaken responses of background clutter by reusing the predicted mask to filter backbone features.
[Paper] [Raw Results] [Training and Testing] [Demo]
Official implementation of the Ocean tracker. Ocean proposes a general anchor-free based tracking framework. It includes a pixel-based anchor-free regression network to solve the weak rectification problem of RPN, and an object-aware classification network to learn robust target-related representation. Moreover, we introduce an effective multi-scale feature combination module to replace heavy result fusion mechanism in recent Siamese trackers. This work also serves as the baseline model of OceanPlus. An additional TensorRT toy demo is provided in this repo.
[Paper] [Raw Results] [Training and Testing] [Demo]
SiamDW is one of the pioneering work using deep backbone networks for Siamese tracking framework. Based on sufficient analysis on network depth, output size, receptive field and padding mode, we propose guidelines to build backbone networks for Siamese tracker. Several deeper and wider networks are built following the guidelines with the proposed CIR module.
-
Tutorial for OceanPlus
Follow OceanPlus [Training and Testing] tutorial
-
Tutorial for Ocean
Follow Ocean [Training and Testing] tutorial
-
Tutorial for SiamDW
Follow SiamDW [Training and Testing] tutorial
experiments:
training and testing settingsdemo:
figures for readmedataset:
testing datasetdata:
training datasetlib:
core scripts for all trackerssnapshot:
pre-trained modelspretrain:
models trained on ImageNet (for training)tutorials:
guidelines for training and testingtracking:
training and testing interface
$TrackSeg
|—— experimnets
|—— lib
|—— snapshot
|—— xxx.model/xxx.pth
|—— dataset
|—— VOT2019.json
|—— VOT2019
|—— ants1...
|—— DAVIS
|—— blackswan...
|—— ...
⚓ Add testing/training code of other trackers.
If any part of our paper or code helps your work, please generouslly cite our work:
@article{OceanPlus_arxiv_2020,
title={Towards Accurate Pixel-wise Object Tracking by Attention Retrieval},
author={Zhipeng Zhang, Bing Li, Weiming Hu, Houwen Peng},
journal={arXiv preprint arXiv:2001.10883},
year={2020}
}
@InProceedings{Ocean_2020_ECCV,
author = {Zhipeng Zhang, Houwen Peng, Jianlong Fu, Bing Li, Weiming Hu},
title = {Ocean: Object-aware Anchor-free Tracking},
booktitle = {The IEEE Conference on European Conference on Computer Vision (ECCV)},
month = {August},
year = {2020}
}
@InProceedings{SiamDW_2019_CVPR,
author = {Zhang, Zhipeng and Peng, Houwen},
title = {Deeper and Wider Siamese Networks for Real-Time Visual Tracking},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2019}
}
@InProceedings{TVOS_2020_CVPR,
author = {Zhang, Yizhuo and Wu, Zhirong and Peng, Houwen and Lin, Stephen},
title = {A Transductive Approach for Video Object Segmentation},
booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2020}
}
[1] Bhat G, Danelljan M, et al. Learning discriminative model prediction for tracking. ICCV2019.
[2] Chen, Kai and Wang, et.al. MMDetection: Open MMLab Detection Toolbox and Benchmark.
[3] Li, B., Wu, W., Wang, Q., et.al. Siamrpn++: Evolution of siamese visual tracking with very deep networks. CVPR2019.
[4] Dai, J., Qi, H., Xiong, Y., et.al. Deformable convolutional networks. ICCV2017.
[5] Wang, Q., Zhang, L., et.al. Fast online object tracking and segmentation: A unifying approach. CVPR2019.
[6] Vu, T., Jang, H., et.al. Cascade RPN: Delving into High-Quality Region Proposal Network with Adaptive Convolution. NIPS2019.
[7] VOT python toolkit: https://github.com/StrangerZhang/pysot-toolkit
🎯 Further discussion anbout our paper and code: zhangzhipeng2017@ia.ac.cn