Deep Regionlets for Object Detection

By Hongyu Xu, Xutao Lv, Xiaoyu Wang, Zhou Ren, Ramalingam Chellappa. 2017

67 citations recorded by the source corpus.

A key challenge in generic object detection is being to handle large variations in object scale, poses, viewpoints, especially part deformations when determining the location for specified object categories. Recent advances in deep neural networks have achieved promising results for object detection by extending the traditional detection methodologies using the convolutional neural network architectures. In this paper, we make an attempt to incorporate another traditional detection schema, Regionlet into an end-to-end trained deep learning framework, and perform ablation studies on its behavior on multiple object detection datasets. More specifically, we propose a "region selection network" and a "gating network". The region selection network serves as a guidance on where to select regions to learn the features from. Additionally, the gating network serves as a local feature selection module to select and transform feature maps to be suitable for detection task. It acts as soft Regionlet selection and pooling. The proposed network is trained end-to-end without additional efforts. Extensive experiments and analysis on the PASCAL VOC dataset and Microsoft COCO dataset show that the proposed framework achieves comparable state-of-the-art results.

Subfields: Computer Vision and Pattern Recognition

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

Citations in the corpus

Bridging the Gap Between Anchor-Based and Anchor-Free Detection via Adaptive Training Sample Selection (2019, neutral): “9 DeepRegionlets [64] trainval35k ResNet-101 39.”
CornerNet: Detecting Objects as Paired Keypoints (2018, neutral): “5 - - - - - Regionlets (Xu et al., 2017) ResNet-101 39.”
Deep High-Resolution Representation Learning for Visual Recognition (2019, neutral): “Deep Regionlets [160] ResNet-101 800 1.”
NISP: Pruning Networks Using Neuron Importance Score Propagation (2017, background): “Despite their impressive predictive power on a wide range of tasks [33, 40, 41, 13, 15, 48, 46, 24, 44, 42, 12, 45], the redundancy in the parameterization of deep learning models has been studied and demonstrated [6].”
Zero-Shot Object Detection (2018, methodology): “Early object detection approaches involved getting object proposals for each image and classifying those object proposals using an image classiﬁcation CNN [16,17,46,55].”
CenterNet: Keypoint Triplets for Object Detection (2019, neutral): “5 - - - - Regionlets [46] ResNet-101 ∼ 1000×600 ∼ 1000×600 39.”
Bottom-Up Object Detection by Grouping Extreme and Center Points (2019, neutral): “5 Deep Regionlets [51] ResNet-101 1000× 600 39.”
Learning Rich Features for Image Manipulation Detection (2018, background): “In contrast to current object detection networks [28, 18, 10, 32, 16, 31] which aim to detect all objects of different categories in an image, a network for image manipulation detection would aim to detect only the tampered regions (usually objects).”
Learning From Noisy Anchors for One-Stage Object Detection (2019, unknown): “2 Regionlets [41] ResNet-101 39.”
High-Resolution Representations for Labeling Pixels and Regions (2019, unknown): “Relation Net [41] ResNet-101 600 - 39:0 58:6 42:9 - - - C-FRCNN [18] ResNet-101 800 1 39:0 59:7 42:8 19:4 42:4 53:0 RetinaNet [63] ResNet-101-FPN 800 1:5 39:1 59:1 42:3 21:8 42:7 50:2 Deep Regionlets [109] ResNet-101 800 1:5 39:3 59:8 - 21:7 43:7 50:9 FitnessNMS [96] ResNet-101 768 - 39:5 58:0 42:6 ”
RepPoints: Point Set Representation for Object Detection (2019, unknown): “2 Deep Regionlets [49] ResNet-101 39.”
Revisiting the Sibling Head in Object Detector (2020, unknown): “5 - - - - Regionlets [37] ResNet-101 39.”
Occlusion-aware R-CNN: Detecting Pedestrians in a Crowd (2018, background): “Recent advances in object detection [20,43,10,27,68,57] are driven by the success of deep convolutional neural networks (CNNs), which uses the bounding box regression techniques to accurately localize the objects based on the deep features.”
Deepstrip: High-Resolution Boundary Refinement (2020, background): “Recent decades have witnessed considerable effort to improve the boundary quality of an object that has been detected [42, 36, 33, 19, 38, 49, 16, 23, 43] or segmented [11, 35, 25, 7, 18].”
CornerNet-Lite: Efficient Keypoint based Object Detection (2019, background): “Other than accuracy [3, 49, 18, 30, 8, 64, 12, 41, 51, 59, 57, 5, 21], many recent works have improved upon the efﬁciency of detectors since the introduction of R-CNN [11], which applies a ConvNet [24] to 2000 RoIs.”