Abstract:Particle filters flexibly represent multiple posterior modes nonparametrically, via a collection of weighted samples, but have classically been applied to tracking problems with known dynamics and observation likelihoods. Such generative models may be inaccurate or unavailable for high-dimensional observations like images. We instead leverage training data to discriminatively learn particle-based representations of uncertainty in latent object states, conditioned on arbitrary observations via deep neural network encoders. While prior discriminative particle filters have used heuristic relaxations of discrete particle resampling, or biased learning by truncating gradients at resampling steps, we achieve unbiased and low-variance gradient estimates by representing posteriors as continuous mixture densities. Our theory and experiments expose dramatic failures of existing reparameterization-based estimators for mixture gradients, an issue we address via an importance-sampling gradient estimator. Unlike standard recurrent neural networks, our mixture density particle filter represents multimodal uncertainty in continuous latent states, improving accuracy and robustness. On a range of challenging tracking and robot localization problems, our approach achieves dramatic improvements in accuracy, while also showing much greater stability across multiple training runs.

Adaptive Probabilistic Tracking with Reliable Particle Selection

Adaptive Tracking Algorithm Based on Particle Grouped Optimization

Enhancing Particle Swarm Optimization Based Particle Filter Tracker

An Efficient Data-Driven Particle PHD Filter for Multitarget Tracking.

Visual Object Tracking Based on Motion-Adaptive Particle Filter under Complex Dynamics

Particle Filter Algorithm Based on Adaptive Resampling Strategy

Embedding Ensemble Tracking in a Stochastic Framework for Robust Object Tracking

Probabilistic visual tracking based on adaptive appearance model

Visual Tracking Via Adaptive Interactive Fusion

Dynamic Target Tracking Based on Particle Filter in Actual Environment

Probabilistic Detection-based Particle Filter for Multi-target Tracking.

Appearance Model-Based Particle Filter for Robust Tracking

Adaptive Target Color Model Updating for Visual Tracking Using Particle Filter

Learning-based Appearance Model for Probabilistic Visual Tracking

Dynamic appearance model for particle filter based visual tracking

Particle Probability Hypothesis Density Filtering for Multitarget Visual Tracking with Robust State Extraction

Differentiable and Stable Long-Range Tracking of Multiple Posterior Modes

An Improved Algorithm for Target Tracking Based on Particle Filter

Interacting multiple model particle filter to adaptive visual tracking.

Probabilistic Multiple Cue Integration for Particle Filter Based Tracking

Learning to track on-the-fly using a particle filter with annealed- weighted QPSO modeled after a singular Dirac delta potential