Point Tracking

46 papers with code • 8 benchmarks • 4 datasets

Point Tracking, often referred to as Tracking any Point (TAP) involves acquiring, focusing on, and continuously tracking specific target point/points across video frames. The system identifies the target point, maintains focus, and predicts its movement, enabling smooth tracking even if the target moves unpredictably, or through occlusions. TAP has wide applications like object tracking, surveillance, and autonomous navigation.

Libraries

Use these libraries to find Point Tracking models and implementations

Most implemented papers

Drag Your GAN: Interactive Point-based Manipulation on the Generative Image Manifold

XingangPan/DragGAN 18 May 2023

Synthesizing visual content that meets users' needs often requires flexible and precise controllability of the pose, shape, expression, and layout of the generated objects.

TAP-Vid: A Benchmark for Tracking Any Point in a Video

deepmind/tapnet 7 Nov 2022

Generic motion understanding from video involves not only tracking objects, but also perceiving how their surfaces deform and move.

TAPIR: Tracking Any Point with per-frame Initialization and temporal Refinement

deepmind/tapnet ICCV 2023

We present a novel model for Tracking Any Point (TAP) that effectively tracks any queried point on any physical surface throughout a video sequence.

PointOdyssey: A Large-Scale Synthetic Dataset for Long-Term Point Tracking

y-zheng18/point_odyssey ICCV 2023

Our goal is to advance the state-of-the-art by placing emphasis on long videos with naturalistic motion.

BootsTAP: Bootstrapped Training for Tracking-Any-Point

google-deepmind/tapnet 1 Feb 2024

To endow models with greater understanding of physics and motion, it is useful to enable them to perceive how solid surfaces move and deform in real scenes.

TAPVid-3D: A Benchmark for Tracking Any Point in 3D

google-deepmind/tapnet 8 Jul 2024

We introduce a new benchmark, TAPVid-3D, for evaluating the task of long-range Tracking Any Point in 3D (TAP-3D).

Beyond Correlation Filters: Learning Continuous Convolution Operators for Visual Tracking

martin-danelljan/Continuous-ConvOp 12 Aug 2016

We also demonstrate the effectiveness of our learning formulation in extensive feature point tracking experiments.

CarFusion: Combining Point Tracking and Part Detection for Dynamic 3D Reconstruction of Vehicles

dineshreddy91/carfusion_to_coco CVPR 2018

In this work, we develop a framework to fuse both the single-view feature tracks and multi-view detected part locations to significantly improve the detection, localization and reconstruction of moving vehicles, even in the presence of strong occlusions.

Muscle Excitation Estimation in Biomechanical Simulation Using NAF Reinforcement Learning

amir-abdi/artisynth_point2point 17 Sep 2018

In this article, we propose a deep reinforcement learning method to estimate the muscle excitations in simulated biomechanical systems.

SD-DefSLAM: Semi-Direct Monocular SLAM for Deformable and Intracorporeal Scenes

UZ-SLAMLab/SD-DefSLAM 19 Oct 2020

Conventional SLAM techniques strongly rely on scene rigidity to solve data association, ignoring dynamic parts of the scene.