The template update problem

By Iain Matthews, Takahiro Ishikawa, Simon Baker. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2004

883 citations recorded by the source corpus.

Subfields: Industrial and Manufacturing Engineering

View Paper: https://www.semanticscholar.org/paper/ef8bc1c3c75b65d0a210f31515818a304dfd22e9

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Industrial and Manufacturing Engineering

Citations in the corpus

Visual tracking via adaptive structural local sparse appearance model (2012, methodology): “Numerous approaches have been proposed for template update [15, 18, 16].”
Understanding and Diagnosing Visual Tracking Systems (2015, neutral): “In [22], Matthews et al.”
Long-term correlation tracking (2015, background): “The first issue is the well-known stability-plasticity dilemma [16, 20].”
Robust tracking using local sparse appearance model and K-selection (2011, methodology): “The generative models formulate the tracking problem as searching for the regions with the highest likelihood [5] [7] [16] [18] [19] [25] [28] [15], either using a single appearance model or multiple appearance models [27].”
On-line selection of discriminative tracking features (2003, methodology): “Anchoring to a reference frame is also used in [22] to avoid drift when updating intensity templates.”
Robust Visual Tracking using 1 Minimization (2009, background): “Approaches have been proposed to overcome the drift problem [25, 21] in different ways.”
Semi-supervised On-Line Boosting for Robust Tracking (2008, background): “Matthews et al. [9] have pinpointed this problem and proposed a partial solu-tion for template trackers.”
Sparse Modeling for Image and Vision Processing (2014, methodology): “We now move to a technique for visual tracking introduced by Mei and Ling [2009], which relies on the sparsity-inducing effect of the `1-norm and on some ideas from the robust face recognition system of Wright et al. [2009]. Given a stream of images x1,x2, .”
Robust Visual Tracking via Convolutional Networks Without Training (2015, methodology): “Although much progress has been made in recent years [1]–[6], it remains a challenging task due to many factors such as illumination changes, partial occlusion, deformation, as well as viewpoint variation [7].”
PointOdyssey: A Large-Scale Synthetic Dataset for Long-Term Point Tracking (2023, methodology): “Our second proposed modification to PIPs aims to tackle this “template update” problem [39].”
In defense of color-based model-free tracking (2015, background): “Such trackers are typically based on templates (e.g. [2] [20] [31]), color representations (e.g. [12] [34] [35]), subspace models (e.g. [9] [38]), and sparse representations (e.g. [6] [24] [30]).”
Robust visual tracking using ℓ1 minimization (2009, background): “Approaches have been proposed to overcome the drift problem [25, 21] in different ways.”
RGB-T Object Tracking: Benchmark and Baseline (2018, neutral): “frame. Let (y t) denote the object descriptor representing a bounding box y t at the t-th frame, and h 0 is learned in the initial frame, which can prevent it from learning drastic appearance changes [62]. Instead of using binary-labeled samples, S-SVM employs structured sample that consists of a ta”
Robust Object Tracking via Sparse Collaborative Appearance Model (2014, background): “This update scheme preserves the first template ψ f (which is usually correct [10], [23], [34]) and takes the most recent tracking result into account.”
Real-Time Tracking via On-line Boosting (2006, methodology): “Furthermore, the on-line trained classifier uses the surrounding background as negative examples in the update and becomes therefore very robust against the drifting problem [18].”
Tracking-Learning-Detection (2012, neutral): “tracking have been proposed [16], [17], [18] as well as”
Brownian descriptor: A rich meta-feature for appearance matching (2014, background): “The prime cause of this drift problem is the change in appearance of object while moving across frames [14, 17].”
Online Object Tracking: A Benchmark (2013, methodology): “[39] address the template update problem for the Lucas-Kanade algorithm [37] where the template is updated with the combination of the fixed reference template extracted from the first frame and the result from the most recent frame.”