Custom Object Detection via Multi-Camera Self-Supervised Learning
Not on the map. Why?
Abstract (arXiv)
This paper proposes MCSSL, a self-supervised learning approach for building custom object detection models in multi-camera networks. MCSSL associates bounding boxes between cameras with overlapping fields of view by leveraging epipolar geometry and state-of-the-art tracking and reID algorithms, and prudently generates two sets of pseudo-labels to fine-tune backbone and detection networks respectively in an object detection model. To train effectively on pseudo-labels,a powerful reID-like pretext task with consistency loss is constructed for model customization. Our evaluation shows that compared with legacy selftraining methods, MCSSL improves average mAP by 5.44% and 6.76% on WildTrack and CityFlow dataset, respectively.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 25 of its 34 references that are in the map. First posted on arXiv 5 Feb 2021.
Builds on
25 of 34 references.
References · 25
- 2015Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networkscited by 9,617
- 2017Mask R-CNNcited by 7,259
- 2018YOLOv3: An Incremental Improvementcited by 2,652
- 2019MMDetection: Open MMLab Detection Toolbox and Benchmarkcited by 1,379
- 2018Domain Adaptive Faster R-CNN for Object Detection in the Wildcited by 442
- 2018Cross-Domain Weakly-Supervised Object Detection through Progressive Domain Adaptationcited by 218
- 2018Using Trusted Data to Train Deep Networks on Labels Corrupted by Severe Noisecited by 216
- 2018SpotTune: Transfer Learning through Adaptive Fine-tuningcited by 166
- 2019CityFlow: A City-Scale Benchmark for Multi-Target Multi-Camera Vehicle Tracking and Re-Identificationcited by 119
- 2019Self-Supervised Learning of 3D Human Pose using Multi-view Geometrycited by 98
- 2018There Are Many Consistent Explanations of Unlabeled Data: Why You Should Averagecited by 82
- 2018M2Det: A Single-Shot Object Detector based on Multi-Level Feature Pyramid Networkcited by 76
13 more
- 2017Camera Style Adaptation for Person Re-identificationcited by 67
- 2016Factors in Finetuning Deep Model for object detectioncited by 53
- 2018Spatial-Temporal Person Re-identificationcited by 45
- 2019CBNet: A Novel Composite Backbone Network Architecture for Object Detectioncited by 44
- 2017Deep Occlusion Reasoning for Multi-Camera Multi-Target Detectioncited by 42
- 2020VehicleNet: Learning Robust Visual Representation for Vehicle Re-identificationcited by 42
- 2017Deep Multi-camera People Detectioncited by 37
- 2018NOTE-RCNN: NOise Tolerant Ensemble RCNN for Semi-Supervised Object Detectioncited by 37
- 2018Procedural Noise Adversarial Examples for Black-Box Attacks on Deep Convolutional Networkscited by 21
- 2019Video Person Re-ID: Fantastic Techniques and Where to Find Themcited by 13
- 2017The WILDTRACK Multi-Camera Person Datasetcited by 10
- 2016Fine-grained Recognition in the Noisy Wild: Sensitivity Analysis of Convolutional Neural Networks Approachescited by 5
- 2019Fast Visual Object Tracking with Rotated Bounding Boxescited by 5
Built on by
Citing papers per year: 2023: 1.
1 paper.