R-FCN: Object Detection via Region-based Fully Convolutional Networks
By Jifeng Dai, Yi Li, Kaiming He. 2016
6,037 citations recorded by the source corpus.
We present region-based, fully convolutional networks for accurate and efficient object detection. In contrast to previous region-based detectors such as Fast/Faster R-CNN that apply a costly per-region subnetwork hundreds of times, our region-based detector is fully convolutional with almost all computation shared on the entire image. To achieve this goal, we propose position-sensitive score maps to address a dilemma between translation-invariance in image classification and translation-variance in object detection. Our method can thus naturally adopt fully convolutional image classifier backbones, such as the latest Residual Networks (ResNets), for object detection. We show competitive results on the PASCAL VOC datasets (e.g., 83.6% mAP on the 2007 set) with the 101-layer ResNet. Meanwhile, our result is achieved at a test-time speed of 170ms per image, 2.5-20 times faster than the Faster R-CNN counterpart. Code is made publicly available at: https://github.com/daijifeng001/r-fcn.
Subfields: Computer Vision and Pattern Recognition, Industrial and Manufacturing Engineering
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
- Luc Van Gool
- Dacheng Tao
- Xiaogang Wang
- Ming-Hsuan Yang
- Trevor Darrell
- Yi Yang
- Jitendra Malik
- Alan L. Yuille
In Industrial and Manufacturing Engineering
- Xiaogang Wang
- Wanli Ouyang
- Bernard Ghanem
- Yancheng Bai
- Iasonas Kokkinos
- Xue Yang
- Junjie Yan
- Yichen Wei
Citations in the corpus
- Mask R-CNN (2017, background): “[26] combined the segment proposal system in [8] and object detection system in [11] for “fully convolutional instance segmentation” (FCIS).”
- Do Better ImageNet Models Transfer Better? (2018, methodology): “In object detection, ImageNet-pretrained networks are used as backbone models for Faster R-CNN and R-FCN detection systems [57, 14].”
- Relation Networks for Object Detection (2017, methodology): “The state-of-the-art object detection methods [24, 18, 38, 9, 32, 10, 23] mostly follow the region based paradigm since it is established in the seminal work R-CNN [19].”
- Do Image Classifiers Generalize Across Time? (2019, methodology): “We use the popular Faster R-CNN [26] and R-FCN [5, 30] architectures for object detection and localization and report results in Table 3.”
- Bridging the Gap Between Anchor-Based and Anchor-Free Detection via Adaptive Training Sample Selection (2019, methodology): “After that, lots of algorithms are proposed to improve its performance, including architecture redesign and reform [4, 9, 5, 28, 30], context and attention mechanism [2, 51, 38, 7, 44], multiscale training and testing [54, 41], training strategy and loss function [40, 52, 61, 17], feature fusion and”
- Cascade R-CNN: High Quality Object Detection and Instance Segmentation (2019, methodology): “To more thoroughly test this claim, a three-stage Cascade R-CNN was implemented with three baseline detectors: Faster R-CNN, R-FCN, and FPN+.”
- Generalized Intersection Over Union: A Metric and a Loss for Bounding Box Regression (2019, background): “Most popular object detectors [20, 21, 3, 12, 13, 16] utilize some combination of the bounding box representations and losses mentioned above.”
- Soft-NMS — Improving Object Detection with One Line of Code (2017, methodology): “To evaluate our method, we experimented with three state-of-the-art detectors, namely, Faster-RCNN [24], R-FCN [16] and Deformable-RFCN.”
- Fully Convolutional Instance-Aware Semantic Segmentation (2016, methodology): “The class-agnostic mask proposals are firstly generated by InstFCN [5], and then classified by R-FCN [9].”
- Pixelwise Instance Segmentation with a Dynamically Instantiated Network (2017, methodology): “We use the publicly available R-FCN object detection framework [12], and ensure that the images used to train the detector do not fall into our test sets for instance segmentation.”
- Towards Accurate Multi-person Pose Estimation in the Wild (2017, methodology): “ system. 3.1. Person Box Detection Our person detector is a Faster-RCNN system [37]. In all experiments reported in this paper we use a ResNet-101 network backbone [22], modified by atrous convolution [9,30] to generate denser feature maps with output stride equal to 8 pixels instead of the default 3”
- Speed/Accuracy Trade-Offs for Modern Convolutional Object Detectors (2016, methodology): “…combinations, as well as some novel ones, and thus identify a set of models which achieve different points on the speed-accuracy tradeoff curve, ranging from fast models, suitable for use on a mobile phone, to a much slower model that achieves a new state of the art on the COCO detection challenge.”
- MegDet: A Large Mini-Batch Object Detector (2017, methodology): “The early generation of CNN-based detectors use very small mini-batch sizes like 2 in Faster-RCNN and R-FCN.”
- DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs (2016, methodology): “Interestingly, the atrous convolution technique has also been adopted for a broader set of tasks, such as object detection [12], [77], instancelevel segmentation [78], visual question answering [79], and optical flow [80].”
- Grounded Language-Image Pre-training (2021, methodology): “Though our current implementation is built upon Dynamic Head (DyHead) [9], our unified formulation can be generalized to any object detection systems [4, 5, 8, 9, 9, 31, 43, 44, 67].”
- iShape: A First Step Towards Irregular Shape Instance Segmentation (2021, neutral): “Compared to object detection [1, 2, 3, 4, 5, 6, 7, 8] and semantic segmen-21 tation [9, 10, 11], instance segmentation provides more fine-grained information but is more challenging and attracts more and more research interests of the community.”
- Deformable Convolutional Networks (2017, methodology): “For Faster RCNN and R-FCN, 256 and 128 RoIs are sampled for the region proposal and the object detection networks, respectively.”