Class balanced underwater object detection dataset generated by class-wise style augmentation

Abstract

Underwater object detection technique is of great significance for various applications in underwater the scenes. However, class imbalance issue is still an unsolved bottleneck for current underwater object detection algorithms. It leads to large precision discrepancies among different classes that the dominant classes with more training data achieve higher detection precisions while the minority classes with fewer training data achieves much lower detection precisions. In this paper, we propose a novel class-wise style augmentation (CWSA) algorithm to generate a class-balanced underwater dataset Balance18 from the public contest underwater dataset URPC2018. CWSA is a new kind of data augmentation technique which augments the training data for the minority classes by generating various colors, textures and contrasts for the minority classes. Compare with previous data augmentation algorithms such flipping, cropping and rotations, CWSA is able to generate a class balanced underwater dataset with diverse color distortions and haze-effects.

References in corpus (4)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• The Effectiveness of Data Augmentation in Image Classification using Deep Learning
• Underwater object detection using Invert Multi-Class Adaboost with deep learning
• A Benchmark dataset for both underwater image enhancement and underwater object detection