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

By Zhaowei Cai, Nuno Vasconcelos. 2019

1,889 citations recorded by the source corpus.

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 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 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.

Subfields: Computer Vision and Pattern Recognition, Education, Sociology and Political Science

How researchers see this paper

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

In Computer Vision and Pattern Recognition

In Education

In Sociology and Political Science

Citations in the corpus

End-to-End Object Detection with Transformers (2020, background): “bility to perform the global computations required for set prediction. 2.3 Object detection Most modern object detection methods make predictions relative to some initial guesses. Two-stage detectors [37,5] predict boxes w.r.t. proposals, whereas single-stage methods make predictions w.r.t. anchors ”
Exploring Plain Vision Transformer Backbones for Object Detection (2022, methodology): “We use the same implementation of Mask R-CNN [24] and Cascade Mask R-CNN [4] for all ViT, Swin, and MViTv2 backbones.”
Deep High-Resolution Representation Learning for Visual Recognition (2019, methodology): “Faster R-CNN [53] Cascade R-CNN [13] FCOS [136] CenterNet [36] R-50 H-18 Table 5 provides the comparison of our method with state-of-the-art methods.”
Image as a Foreign Language: BEiT Pretraining for All Vision and Vision-Language Tasks (2022, methodology): “BEIT-3 uses Cascade Mask R-CNN [CV21] as the detection head.”
Seesaw Loss for Long-Tailed Instance Segmentation (2020, methodology): “We implement our method with mmdetection [5] and train Mask R-CNN [17], Cascade Mask R-CNN [1] using the 2x training schedule [5, 13].”
CLOCs: Camera-LiDAR Object Candidates Fusion for 3D Object Detection (2020, background): “[29], MS-CNN [30] and Cascade R-CNN [31].”
Image as a Foreign Language: BEIT Pretraining for Vision and Vision-Language Tasks (2023, methodology): “BEIT-3 uses Cascade Mask R-CNN [7] as the detection head.”
BEVDet: High-performance Multi-camera 3D Object Detection in Bird-Eye-View (2021, methodology): “Two-stage method FasterRCNN [33], one-stage method RetinaNet [20], and their derivatives [10, 2, 4, 39, 49] are the dominant methods in this area [21, 35].”
A survey of the recent architectures of deep convolutional neural networks (2019, background): “, modified the architectures as mentioned earlier to improve the performance (Lin et al. 2017; Cai and Vasconcelos 2019; Pang et al. 2020).”
Pedestrian Detection: The Elephant In The Room (2020, methodology): “Since most of the top ranked methods on Caltech, CityPersons and ECP are direct extension of Faster/Mask R-CNN [30,16] family, we also select recent Cascade R-CNN [8] (an extension of R-CNN family) as our baseline.”
DilateFormer: Multi-Scale Dilated Transformer for Visual Recognition (2023, methodology): “We utilize two representative frameworks: Mask R-CNN [12] and Cascade Mask R-CNN [108] implemented in mmdetection[109]andadopttheImageNet-1Kpre-trainedvari-ants as backbones.”
ViTAEv2: Vision Transformer Advanced by Exploring Inductive Bias for Image Recognition and Beyond (2022, neutral): “4 82 (Cai and Vasconcelos, 2019) Swin-T (Liu et al, 2021) ICCV’21 50.”
InternImage: Exploring Large-Scale Vision Foundation Models with Deformable Convolutions (2022, methodology): “With the 3 × multi-scale training schedule, more parameters, and more advanced Cascade Mask R-CNN [67], InternImage-XL achieves AP b of 56.2, surpassing ConvNeXt-XL by 1.0 points (56.2 vs .”