Augmentation for small object detection

Published in

International Conference on Advances in Computi...(2019)

External Links:

Generate Graph DownloadPDF

TL;DR

This work analyzes the current state-of-the-art model, Mask-RCNN, on a challenging dataset, MS COCO, and shows that the overlap between small ground-truth objects and the predicted anchors is much lower than the expected IoU threshold.

Abstract

In recent years, object detection has experienced impressive progress. Despite these improvements, there is still a significant gap in the performance between the detection of small and large objects. We analyze the current state-of-the-art model, Mask-RCNN, on a challenging dataset, MS COCO. We show that the overlap between small ground-truth objects and the predicted anchors is much lower than the expected IoU threshold. We conjecture this is due to two factors; (1) only a few images are containing small objects, and (2) small objects do not appear enough even within each image containing them. We thus propose to oversample those images with small objects and augment each of those images by copy-pasting small objects many times. It allows us to trade off the quality of the detector on large objects with that on small objects. We evaluate different pasting augmentation strategies, and ultimately, we achieve 9.7\% relative improvement on the instance segmentation and 7.1\% on the object detection of small objects, compared to the current state of the art method on MS COCO.

Authors

Z. Wojna

4 papers

Kyunghyun Cho

23 papers

Mate Kisantal

1 papers

References36 items

Small-objectness sensitive detection based on shifted single shot detector

Small Object Detection in Optical Remote Sensing Images via Modified Faster R-CNN

LOCO: Local Context Based Faster R-CNN for Small Traffic Sign Detection

CNN-based small object detection and visualization with feature activation mapping

A systematic study of the class imbalance problem in convolutional neural networks

Augmentation for small object detection

Published in

International Conference on Advances in Computi...(2019)

External Links:

Generate Graph DownloadPDF

TL;DR

Abstract

Authors

Z. Wojna

4 papers

Kyunghyun Cho

23 papers

Mate Kisantal

1 papers

References36 items

Small-objectness sensitive detection based on shifted single shot detector

Small Object Detection in Optical Remote Sensing Images via Modified Faster R-CNN

LOCO: Local Context Based Faster R-CNN for Small Traffic Sign Detection

CNN-based small object detection and visualization with feature activation mapping

A systematic study of the class imbalance problem in convolutional neural networks

Jakub Murawski

1 papers

Jacek Naruniec

2 papers

Feature-fused SSD: fast detection for small objects

The Devil is in the Decoder: Classification, Regression and GANs

Perceptual Generative Adversarial Networks for Small Object Detection

Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour

Improving Small Object Proposals for Company Logo Detection

Semantic Instance Segmentation via Deep Metric Learning

DSSD : Deconvolutional Single Shot Detector

Finding Tiny Faces

Feature Pyramid Networks for Object Detection

Speed/Accuracy Trade-Offs for Modern Convolutional Object Detectors

Fully Convolutional Instance-Aware Semantic Segmentation

R-CNN for Small Object Detection

Exploit All the Layers: Fast and Accurate CNN Object Detector with Scale Dependent Pooling and Cascaded Rejection Classifiers

Semantic Segmentation of Small Objects and Modeling of Uncertainty in Urban Remote Sensing Images Using Deep Convolutional Neural Networks

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

Instance-Aware Semantic Segmentation via Multi-task Network Cascades

Inside-Outside Net: Detecting Objects in Context with Skip Pooling and Recurrent Neural Networks

SSD: Single Shot MultiBox Detector

3D Object Proposals for Accurate Object Class Detection

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

Microsoft COCO: Common Objects in Context

Real-Time Traffic Sign Recognition System based on Colour Image Segmentation

Joint Deep Learning for Pedestrian Detection

Classification of small lesions in dynamic breast MRI: eliminating the need for precise lesion segmentation through spatio-temporal analysis of contrast enhancement

Traffic sign recognition with multi-scale Convolutional Networks

Small object recognition techniques based on structured template matching for high-resolution satellite images

Automated vision system for localizing structural defects in textile fabrics

Automatic thresholding for defect detection

Detection and classification of lobular and DCIS (small cell) microcalcifications in digital mammograms

Field of Study

Computer Science

Journal Information

Name

ArXiv

Volume

abs/2005.00687

Venue Information

Name

International Conference on Advances in Computing and Information Technology

Type

conference

URL

N/A

Alternate Names

ACITY
Int Conf Adv Comput Inf Technol

TL;DR

Abstract

Authors

References36 items

Small-objectness sensitive detection based on shifted single shot detector

Small Object Detection in Optical Remote Sensing Images via Modified Faster R-CNN

LOCO: Local Context Based Faster R-CNN for Small Traffic Sign Detection

CNN-based small object detection and visualization with feature activation mapping

A systematic study of the class imbalance problem in convolutional neural networks

TL;DR

Abstract

Authors

References36 items

Small-objectness sensitive detection based on shifted single shot detector

Small Object Detection in Optical Remote Sensing Images via Modified Faster R-CNN

LOCO: Local Context Based Faster R-CNN for Small Traffic Sign Detection

CNN-based small object detection and visualization with feature activation mapping

A systematic study of the class imbalance problem in convolutional neural networks

Feature-fused SSD: fast detection for small objects

The Devil is in the Decoder: Classification, Regression and GANs

Perceptual Generative Adversarial Networks for Small Object Detection

Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour

Improving Small Object Proposals for Company Logo Detection

Semantic Instance Segmentation via Deep Metric Learning

Mask R-CNN

DSSD : Deconvolutional Single Shot Detector

Finding Tiny Faces

Feature Pyramid Networks for Object Detection

Speed/Accuracy Trade-Offs for Modern Convolutional Object Detectors

Fully Convolutional Instance-Aware Semantic Segmentation

R-CNN for Small Object Detection

Exploit All the Layers: Fast and Accurate CNN Object Detector with Scale Dependent Pooling and Cascaded Rejection Classifiers

Semantic Segmentation of Small Objects and Modeling of Uncertainty in Urban Remote Sensing Images Using Deep Convolutional Neural Networks

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

Instance-Aware Semantic Segmentation via Multi-task Network Cascades

Inside-Outside Net: Detecting Objects in Context with Skip Pooling and Recurrent Neural Networks

SSD: Single Shot MultiBox Detector

3D Object Proposals for Accurate Object Class Detection

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

Microsoft COCO: Common Objects in Context

Real-Time Traffic Sign Recognition System based on Colour Image Segmentation

Joint Deep Learning for Pedestrian Detection

Classification of small lesions in dynamic breast MRI: eliminating the need for precise lesion segmentation through spatio-temporal analysis of contrast enhancement

Traffic sign recognition with multi-scale Convolutional Networks

Small object recognition techniques based on structured template matching for high-resolution satellite images

Automated vision system for localizing structural defects in textile fabrics

Automatic thresholding for defect detection

Detection and classification of lobular and DCIS (small cell) microcalcifications in digital mammograms

Detectron

Field of Study

Journal Information

Name

Volume

Venue Information

Name

Type

URL

Alternate Names