Cascade R-CNN: High Quality Object Detection and Instance Segmentation

Abstract

In object detection, the intersection over union (IoU) threshold is frequently used to define positives/negatives. The threshold used to train a detector defines its \textit{quality}. While the commonly used threshold of 0.5 leads to noisy (low-quality) detections, detection performance frequently degrades for larger thresholds. This paradox of high-quality detection has two causes: 1) overfitting, due to vanishing positive samples for large thresholds, and 2) inference-time quality mismatch between detector and test hypotheses. A multi-stage object detection architecture, the Cascade R-CNN, composed of a sequence of detectors trained with increasing IoU thresholds, is proposed to address these problems. The detectors are trained sequentially, using the output of a detector as training set for the next. This resampling progressively improves hypotheses quality, guaranteeing a positive training set of equivalent size for all detectors and minimizing overfitting. The same cascade is applied at inference, to eliminate quality mismatches between hypotheses and detectors. An implementation of the Cascade R-CNN without bells or whistles achieves state-of-the-art performance on the COCO dataset, and significantly improves high-quality detection on generic and specific object detection datasets, including VOC, KITTI, CityPerson, and WiderFace. Finally, the Cascade R-CNN is generalized to instance segmentation, with nontrivial improvements over the Mask R-CNN. To facilitate future research, two implementations are made available at \url{https://github.com/zhaoweicai/cascade-rcnn} (Caffe) and \url{https://github.com/zhaoweicai/Detectron-Cascade-RCNN} (Detectron).

extension of arXiv:1712.00726 "Cascade R-CNN: Delving into High Quality Object Detection"

References in corpus (8)

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
• Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition
• Hybrid Task Cascade for Instance Segmentation
• Speed/accuracy trade-offs for modern convolutional object detectors
• Rethinking ImageNet Pre-training
• Learning Chained Deep Features and Classifiers for Cascade in Object Detection
• Single-Shot Bidirectional Pyramid Networks for High-Quality Object Detection

Cited by in corpus (17)

• Deep High-Resolution Representation Learning for Visual Recognition
• SkyNet: a Hardware-Efficient Method for Object Detection and Tracking on Embedded Systems
• Detection in Crowded Scenes: One Proposal, Multiple Predictions
• Boundary-preserving Mask R-CNN
• Cascaded Human-Object Interaction Recognition
• Learning from Noisy Anchors for One-stage Object Detection
• Semantic Head Enhanced Pedestrian Detection in a Crowd
• Seesaw Loss for Long-Tailed Instance Segmentation
• PSC-Net: Learning Part Spatial Co-occurrence for Occluded Pedestrian Detection
• Dually Supervised Feature Pyramid for Object Detection and Segmentation
• AutoLoss-Zero: Searching Loss Functions from Scratch for Generic Tasks
• A Selective Survey on Versatile Knowledge Distillation Paradigm for Neural Network Models
• Query-based Hard-Image Retrieval for Object Detection at Test Time
• Towards Early Prediction of Human iPSC Reprogramming Success
• Neighbor-Vote: Improving Monocular 3D Object Detection through Neighbor Distance Voting
• Target-Aware Object Discovery and Association for Unsupervised Video Multi-Object Segmentation
• Survey on Deep Learning-based Kuzushiji Recognition