Small Object Detection via Pixel Level Balancing With Applications to Blood Cell Detection (2022-06-17T00:00:00.000000Z)

TL;DR

A method called pixel level balancing (PLB), which takes into account the number of pixels contained in the detection box as an impact factor to characterize the size of the inspected objects, and uses this as animpact factor to improve the accuracy of small object detection.

Abstract

Object detection technology has been widely used in medical field, such as detecting the images of blood cell to count the changes and distribution for assisting the diagnosis of diseases. However, detecting small objects is one of the most challenging and important problems especially in medical scenarios. Most of the objects in medical images are very small but influential. Improving the detection performance of small objects is a very meaningful topic for medical detection. Current researches mainly focus on the extraction of small object features and data augmentation for small object samples, all of these researches focus on extracting the feature space of small objects better. However, in the training process of a detection model, objects of different sizes are mixed together, which may interfere with each other and affect the performance of small object detection. In this paper, we propose a method called pixel level balancing (PLB), which takes into account the number of pixels contained in the detection box as an impact factor to characterize the size of the inspected objects, and uses this as an impact factor. The training loss of each object of different size is adjusted by a weight dynamically, so as to improve the accuracy of small object detection. Finally, through experiments, we demonstrate that the size of objects in object detection interfere with each other. So that we can improve the accuracy of small object detection through PLB operation. This method can perform well with blood cell detection in our experiments.

Authors

Minglei Tong

2 papers

Pengzhi Chu

2 papers

B. Hu

1 papers

TL;DR

Abstract

Authors

References68 items

Structure Guided Deep Neural Network for Unsupervised Active Learning

DDTNet: A dense dual-task network for tumor-infiltrating lymphocyte detection and segmentation in histopathological images of breast cancer

The KFIoU Loss for Rotated Object Detection

Deep Unsupervised Active Learning via Matrix Sketching

Learning High-Precision Bounding Box for Rotated Object Detection via Kullback-Leibler Divergence

Rethinking Rotated Object Detection with Gaussian Wasserstein Distance Loss

Dense Label Encoding for Boundary Discontinuity Free Rotation Detection

Image Matching from Handcrafted to Deep Features: A Survey

An Automatic Nucleus Segmentation and CNN Model based Classification Method of White Blood Cell

Learning for Graph Matching and Related Combinatorial Optimization Problems

Combinatorial Learning of Robust Deep Graph Matching: An Embedding Based Approach

Identity-aware CycleGAN for face photo-sketch synthesis and recognition

SCRDet++: Detecting Small, Cluttered and Rotated Objects via Instance-Level Feature Denoising and Rotation Loss Smoothing

Unifying Offline and Online Multi-Graph Matching via Finding Shortest Paths on Supergraph

YOLOv4: Optimal Speed and Accuracy of Object Detection

On the Arbitrary-Oriented Object Detection: Classification Based Approaches Revisited

Arbitrary-Oriented Object Detection with Circular Smooth Label

Neural Graph Matching Network: Learning Lawler’s Quadratic Assignment Problem With Extension to Hypergraph and Multiple-Graph Matching

EfficientDet: Scalable and Efficient Object Detection

Learning Modulated Loss for Rotated Object Detection

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

EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks

Augmentation for small object detection

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

YOLOv3: An Incremental Improvement

Adaptive Discrete Hypergraph Matching

Cell detection in pathology and microscopy images with multi-scale fully convolutional neural networks

Understanding the Disharmony Between Dropout and Batch Normalization by Variance Shift

FSSD: Feature Fusion Single Shot Multibox Detector

Faster R-CNN based microscopic cell detection

Squeeze-and-Excitation Networks

Focal Loss for Dense Object Detection

DSOD: Learning Deeply Supervised Object Detectors from Scratch

Enhancement of SSD by concatenating feature maps for object detection

A survey on deep learning in medical image analysis

DSSD : Deconvolutional Single Shot Detector

YOLO9000: Better, Faster, Stronger

Beyond Skip Connections: Top-Down Modulation for Object Detection

Feature Pyramid Networks for Object Detection

Densely Connected Convolutional Networks

Lung nodule detection in CT images using deep convolutional neural networks

A Short Survey of Recent Advances in Graph Matching

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

Deep learning for digital pathology image analysis: A comprehensive tutorial with selected use cases

Deep Residual Learning for Image Recognition

SSD: Single Shot MultiBox Detector

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

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

Improving Computer-Aided Detection Using Convolutional Neural Networks and Random View Aggregation

Fast R-CNN

Joint Active Learning with Feature Selection via CUR Matrix Decomposition

Multi-Graph Matching via Affinity Optimization with Graduated Consistency Regularization

Consistency-Driven Alternating Optimization for Multigraph Matching: A Unified Approach

Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition

Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation

Mitosis Detection in Breast Cancer Histology Images with Deep Neural Networks

Image Fusion Based on Nonsubsampled Contourlet Transform and Saliency-Motivated Pulse Coupled Neural Networks

Region-based multi-focus image fusion using the local spatial frequency

Efficient Non-Maximum Suppression

Deformable models in medical image analysis

Deformable models in medical image analysis: a survey

Rich feature hierarchies for accurate object detection and semantic segmentation

Cell Detection and Segmentation in Microscopy Images with Improved Mask R-CNN

Ef fi cient Non - maximum Suppression . ” in 18 th 457 International Conference on Pattern Recognition ( ICPR ’ 06 ) , Hong Kong , China , August , 2006

Send Orders of Reprints at Reprints@benthamscience.net Viewpoints on Medical Image Processing: from Science to Application

Instead of setting ﬁ xed empirical values before training

With our proposed method PLB, the training trend of two-stage detector can be adjusted and the detection accuracy of small objects can be further improved

Con ﬂ ict of Interest: Author QK was employed by Riseye Intelligent Technology (

Field of Study

Journal Information

Name

Volume

Venue Information

Name

Type

URL