Context-aware Deep Feature Compression for High-speed Visual Tracking

Abstract

We propose a new context-aware correlation filter based tracking framework to achieve both high computational speed and state-of-the-art performance among real-time trackers. The major contribution to the high computational speed lies in the proposed deep feature compression that is achieved by a context-aware scheme utilizing multiple expert auto-encoders; a context in our framework refers to the coarse category of the tracking target according to appearance patterns. In the pre-training phase, one expert auto-encoder is trained per category. In the tracking phase, the best expert auto-encoder is selected for a given target, and only this auto-encoder is used. To achieve high tracking performance with the compressed feature map, we introduce extrinsic denoising processes and a new orthogonality loss term for pre-training and fine-tuning of the expert auto-encoders. We validate the proposed context-aware framework through a number of experiments, where our method achieves a comparable performance to state-of-the-art trackers which cannot run in real-time, while running at a significantly fast speed of over 100 fps.

9 pages, 6 figures, Accepted in CVPR2018 (IEEE conference on Computer Vision and Pattern Recognition)

References in corpus (8)

• High-Speed Tracking with Kernelized Correlation Filters
• Beyond Correlation Filters: Learning Continuous Convolution Operators for Visual Tracking
• Learning Spatially Regularized Correlation Filters for Visual Tracking
• Return of the Devil in the Details: Delving Deep into Convolutional Nets
• MatConvNet - Convolutional Neural Networks for MATLAB
• Siamese Instance Search for Tracking
• Switching Convolutional Neural Network for Crowd Counting
• Context-aware CNNs for person head detection

Cited by in corpus (32)

• Deep Learning for Visual Tracking: A Comprehensive Survey
• Distilled Siamese Networks for Visual Tracking
• Learning Reinforced Attentional Representation for End-to-End Visual Tracking
• Pedestrian Models for Autonomous Driving Part I: Low-Level Models, from Sensing to Tracking
• Event-based visual place recognition with ensembles of temporal windows
• One-Shot Recognition of Manufacturing Defects in Steel Surfaces
• Real-Time Correlation Tracking via Joint Model Compression and Transfer
• Accurate Bounding-box Regression with Distance-IoU Loss for Visual Tracking
• Target-Aware Deep Tracking
• Siamese Cascaded Region Proposal Networks for Real-Time Visual Tracking
• LaSOT: A High-quality Benchmark for Large-scale Single Object Tracking
• Learning the Model Update for Siamese Trackers
• LaSOT: A High-quality Large-scale Single Object Tracking Benchmark
• Remove Cosine Window from Correlation Filter-based Visual Trackers: When and How
• Detection and Tracking Meet Drones Challenge
• Siamese Keypoint Prediction Network for Visual Object Tracking
• A Review of Visual Trackers and Analysis of its Application to Mobile Robot
• Unsupervised Deep Representation Learning for Real-Time Tracking
• Real-Time Object Tracking via Meta-Learning: Efficient Model Adaptation and One-Shot Channel Pruning
• Online Visual Tracking with One-Shot Context-Aware Domain Adaptation
• Pose-Assisted Multi-Camera Collaboration for Active Object Tracking
• Hard Negative Samples Emphasis Tracker without Anchors
• Hallucinated Adversarial Learning for Robust Visual Tracking
• ParaDiS: Parallelly Distributable Slimmable Neural Networks
• Rethinking Convolutional Features in Correlation Filter Based Tracking
• Multiple Convolutional Features in Siamese Networks for Object Tracking
• Implications of Human Irrationality for Reinforcement Learning
• Material Based Object Tracking in Hyperspectral Videos: Benchmark and Algorithms
• Guided Feature Selection for Deep Visual Odometry
• Object Tracking in Hyperspectral Videos with Convolutional Features and Kernelized Correlation Filter
• Object-Adaptive LSTM Network for Real-time Visual Tracking with Adversarial Data Augmentation
• Multi-Branch Siamese Networks with Online Selection for Object Tracking