SCRDet++: Detecting Small, Cluttered and Rotated Objects via Instance-Level Feature Denoising and Rotation Loss Smoothing (2020-04-28T00:00:00.000000Z)

TL;DR

Instance-level denoising on the feature map is performed to enhance the detection to small and cluttered objects and a novel IoU constant factor is added to the smooth L1 loss to address the long standing boundary problem.

Abstract

Small and cluttered objects are common in real-world which are challenging for detection. The difficulty is further pronounced when the objects are rotated, as traditional detectors often routinely locate the objects in horizontal bounding box such that the region of interest is contaminated with background or nearby interleaved objects. In this paper, we first innovatively introduce the idea of denoising to object detection. Instance-level denoising on the feature map is performed to enhance the detection to small and cluttered objects. To handle the rotation variation, we also add a novel IoU constant factor to the smooth L1 loss to address the long standing boundary problem, which to our analysis, is mainly caused by the periodicity of angular (PoA) and exchangeability of edges (EoE). By combing these two features, our proposed detector is termed as SCRDet++. Extensive experiments are performed on large aerial images public datasets DOTA, DIOR, UCAS-AOD as well as natural image dataset COCO, scene text dataset ICDAR2015, small traffic light dataset BSTLD and our released S <inline-formula><tex-math notation="LaTeX">$^{2}$</tex-math><alternatives><mml:math><mml:msup><mml:mrow/><mml:mn>2</mml:mn></mml:msup></mml:math><inline-graphic xlink:href="yan-ieq1-3166956.gif"/></alternatives></inline-formula> TLD by this paper. The results show the effectiveness of our approach. The released dataset S <inline-formula><tex-math notation="LaTeX">$^{2}$</tex-math><alternatives><mml:math><mml:msup><mml:mrow/><mml:mn>2</mml:mn></mml:msup></mml:math><inline-graphic xlink:href="yan-ieq2-3166956.gif"/></alternatives></inline-formula> TLD is made public available, which contains 5,786 images with 14,130 traffic light instances across five categories.

References88 items

Cross-Scale Feature Fusion for Object Detection in Optical Remote Sensing Images

Arbitrary-Oriented Object Detection in Remote Sensing Images Based on Polar Coordinates

HAMBox: Delving Into Mining High-Quality Anchors on Face Detection

FMSSD: Feature-Merged Single-Shot Detection for Multiscale Objects in Large-Scale Remote Sensing Imagery

Extended Feature Pyramid Network for Small Object Detection

Axis Learning for Orientated Objects Detection in Aerial Images

Arbitrary-Oriented Object Detection with Circular Smooth Label

Multi-Scale Feature Integrated Attention-Based Rotation Network for Object Detection in VHR Aerial Images

Rotation-aware and multi-scale convolutional neural network for object detection in remote sensing images

RADet: Refine Feature Pyramid Network and Multi-Layer Attention Network for Arbitrary-Oriented Object Detection of Remote Sensing Images

Objects detection for remote sensing images based on polar coordinates

Deep Learning on Image Denoising: An overview

Oriented Objects as pairs of Middle Lines

HAMBox: Delving into Online High-quality Anchors Mining for Detecting Outer Faces

Mask OBB: A Semantic Attention-Based Mask Oriented Bounding Box Representation for Multi-Category Object Detection in Aerial Images

IENet: Interacting Embranchment One Stage Anchor Free Detector for Orientation Aerial Object Detection

Gliding Vertex on the Horizontal Bounding Box for Multi-Oriented Object Detection

Learning Modulated Loss for Rotated Object Detection

Enhanced Feature Representation in Detection for Optical Remote Sensing Images

Feature-Attentioned Object Detection in Remote Sensing Imagery

Object Detection in Optical Remote Sensing Images: A Survey and A New Benchmark

R3Det: Refined Single-Stage Detector with Feature Refinement for Rotating Object

DAPAS : Denoising Autoencoder to Prevent Adversarial attack in Semantic Segmentation

A2RMNet: Adaptively Aspect Ratio Multi-Scale Network for Object Detection in Remote Sensing Images

Adaptive Period Embedding for Representing Oriented Objects in Aerial Images

Learning RoI Transformer for Oriented Object Detection in Aerial Images

CenterNet: Keypoint Triplets for Object Detection

FCOS: Fully Convolutional One-Stage Object Detection

CAD-Net: A Context-Aware Detection Network for Objects in Remote Sensing Imagery

Augmentation for small object detection

IoU-Adaptive Deformable R-CNN: Make Full Use of IoU for Multi-Class Object Detection in Remote Sensing Imagery

Bottom-Up Object Detection by Grouping Extreme and Center Points

Feature Denoising for Improving Adversarial Robustness

SCRDet: Towards More Robust Detection for Small, Cluttered and Rotated Objects

Salience Biased Loss for Object Detection in Aerial Images

Deformable Faster R-CNN with Aggregating Multi-Layer Features for Partially Occluded Object Detection in Optical Remote Sensing Images

CornerNet: Detecting Objects as Paired Keypoints

Towards Multi-class Object Detection in Unconstrained Remote Sensing Imagery

Position Detection and Direction Prediction for Arbitrary-Oriented Ships via Multitask Rotation Region Convolutional Neural Network

SNIPER: Efficient Multi-Scale Training

YOLOv3: An Incremental Improvement

Rotation-Sensitive Regression for Oriented Scene Text Detection

Path Aggregation Network for Instance Segmentation

Seeing Small Faces from Robust Anchor's Perspective

Automatic Ship Detection in Remote Sensing Images from Google Earth of Complex Scenes Based on Multiscale Rotation Dense Feature Pyramid Networks

TextBoxes++: A Single-Shot Oriented Scene Text Detector

FOTS: Fast Oriented Text Spotting with a Unified Network

Deformable ConvNet with Aspect Ratio Constrained NMS for Object Detection in Remote Sensing Imagery

DOTA: A Large-Scale Dataset for Object Detection in Aerial Images

Learning a Rotation Invariant Detector with Rotatable Bounding Box

Non-local Neural Networks

Squeeze-and-Excitation Networks

Focal Loss for Dense Object Detection

R2CNN: Rotational Region CNN for Orientation Robust Scene Text Detection

An Efficient and Robust Integrated Geospatial Object Detection Framework for High Spatial Resolution Remote Sensing Imagery

Perceptual Generative Adversarial Networks for Small Object Detection

A deep learning approach to traffic lights: Detection, tracking, and classification

EAST: An Efficient and Accurate Scene Text Detector

Mask R-CNN

Deformable Convolutional Networks

Arbitrary-Oriented Scene Text Detection via Rotation Proposals

A High Resolution Optical Satellite Image Dataset for Ship Recognition and Some New Baselines

DSSD : Deconvolutional Single Shot Detector

YOLO9000: Better, Faster, Stronger

Feature Pyramid Networks for Object Detection

Aggregated Residual Transformations for Deep Neural Networks

Learning Rotation-Invariant Convolutional Neural Networks for Object Detection in VHR Optical Remote Sensing Images

R-FCN: Object Detection via Region-based Fully Convolutional Networks

Stacked Hourglass Networks for Human Pose Estimation

Deep Residual Learning for Image Recognition

Orientation robust object detection in aerial images using deep convolutional neural network

SSD: Single Shot MultiBox Detector

Multi-Scale Context Aggregation by Dilated Convolutions

ICDAR 2015 competition on Robust Reading

You Only Look Once: Unified, Real-Time Object Detection

Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks

Fast R-CNN

Very Deep Convolutional Networks for Large-Scale Image Recognition

Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition

Microsoft COCO: Common Objects in Context

Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation

Adaptive denoising filtering for object detection applications

SARD: Towards Scale-Aware Rotated Object Detection in Aerial Imagery

Single Shot Anchor Refinement Network for Oriented Object Detection in Optical Remote Sensing Imagery

International Journal of Computer Vision manuscript No. (will be inserted by the editor) The PASCAL Visual Object Classes (VOC) Challenge

Ã Indicates our own implementation, higher than the official baseline. y Indicates data augmentation is used

PhD candidate with Department of Computer Science and Engineering, Shanghai Jiao Tong University, with research interests in robotics

A Deep Learning Approach to Trafﬁc Lights: Detection, Tracking, and Classiﬁcation