RPT: Learning Point Set Representation for Siamese Visual Tracking[ECCVW2020]

:sunny: Currently, this code only supports the offline version of RPT.

News

• :trophy: We are the Winner of VOT-2020 Short-Term challenge
• :trophy: We get both 1st on the public and sequestered benchmark dataset of the VOT2020 Short-Term challenge
• :sunny::sunny:Our VOT2020-ST Winner presentation has been uploaded

Spotlight video

Models

• The pretrained model can be downloaded from [google] or [baidu], extraction code: g4ac.
• The raw results can be downloaded [here], extraction code: mkbh.

Abstract

While remarkable progress has been made in robust visual tracking, accurate target state estimation still remains a highly challenging problem. In this paper, we argue that this issue is closely related to the prevalent bounding box representation, which provides only a coarse spatial extent of object. Thus an effcient visual tracking framework is proposed to accurately estimate the target state with a finer representation as a set of representative points. The point set is trained to indicate the semantically and geometrically significant positions of target region, enabling more fine-grained localization and modeling of object appearance. We further propose a multi-level aggregation strategy to obtain detailed structure information by fusing hierarchical convolution layers. Extensive experiments on several challenging benchmarks including OTB2015, VOT2018, VOT2019 and GOT-10k demonstrate that our method achieves new state-of-the-art performance while running at over 20 FPS.

Installation

Please find installation instructions in INSTALL.md

Quick Start: Using siamreppoints

Download pretrained models and put the siamreppoints.model in the correct directory in experiments

cd siamreppoints/tools
python test.py \
       --snapshot ./snapshot/siamreppoints.model \  #model path
       --dataset VOT2018 \  #dataset name
       --config ./experiments/siamreppoints/config_vot2018_offline.yaml  #config file 

cd siamreppoints/tools
python eval.py \
       --tracker_path ./results \  #result path
       --dataset VOT2018 \  #dataset name
       --tracker_prefix 'siam' \  # tracker_name
       --num 1  # number thread to eval

Ackowledgement

• pysot