Deep Feature Augmentation for Occluded Image Classification

Abstract

Due to the difficulty in acquiring massive task-specific occluded images, the classification of occluded images with deep convolutional neural networks (CNNs) remains highly challenging. To alleviate the dependency on large-scale occluded image datasets, we propose a novel approach to improve the classification accuracy of occluded images by fine-tuning the pre-trained models with a set of augmented deep feature vectors (DFVs). The set of augmented DFVs is composed of original DFVs and pseudo-DFVs. The pseudo-DFVs are generated by randomly adding difference vectors (DVs), extracted from a small set of clean and occluded image pairs, to the real DFVs. In the fine-tuning, the back-propagation is conducted on the DFV data flow to update the network parameters. The experiments on various datasets and network structures show that the deep feature augmentation significantly improves the classification accuracy of occluded images without a noticeable influence on the performance of clean images. Specifically, on the ILSVRC2012 dataset with synthetic occluded images, the proposed approach achieves 11.21% and 9.14% average increases in classification accuracy for the ResNet50 networks fine-tuned on the occlusion-exclusive and occlusion-inclusive training sets, respectively.

References in corpus (5)

• Data Augmentation Generative Adversarial Networks
• Generative Image Inpainting with Contextual Attention
• MS-Celeb-1M: A Dataset and Benchmark for Large-Scale Face Recognition
• Occlusion Robust Face Recognition Based on Mask Learning with PairwiseDifferential Siamese Network
• Boosting Occluded Image Classification via Subspace Decomposition Based Estimation of Deep Features

Cited by in corpus (11)

• A Comprehensive Survey of Image Augmentation Techniques for Deep Learning
• Momentum Contrastive Learning for Few-Shot COVID-19 Diagnosis from Chest CT Images
• Dynamic Label Assignment for Object Detection by Combining Predicted IoUs and Anchor IoUs
• An Unsupervised Domain Adaptation Model based on Dual-module Adversarial Training
• Enhanced U-Net: A Feature Enhancement Network for Polyp Segmentation
• Colonoscopy Polyp Detection and Classification: Dataset Creation and Comparative Evaluations
• SGNet: A Super-class Guided Network for Image Classification and Object Detection
• Few-Shot Learning by Integrating Spatial and Frequency Representation
• Audio-Visual Transformer Based Crowd Counting
• Carton dataset synthesis method for domain shift based on foreground texture decoupling and replacement
• Training Deep Neural Networks via Branch-and-Bound