RADAM: Texture recognition through randomized aggregated encoding of deep activation maps

Author(s)

Scabini, Leonardo ; Zielinski, Kallil M. ; Ribas, Lucas C. ; Goncalves, Wesley N. ; De Baets, Bernard ; Bruno, Odemir M.

• Document type: Journal article
• Source: PATTERN RECOGNITION; v. 143, p. 13-pg., 2023-11-01.

Abstract

Texture analysis is a classical yet challenging task in computer vision for which deep neural networks are actively being applied. Most approaches are based on building feature aggregation modules around a pre-trained backbone and then fine-tuning the new architecture on specific texture recognition tasks. Here we propose a new method named Random encoding of Aggregated Deep Activation Maps (RADAM) which extracts rich texture representations without ever changing the backbone. The technique consists of encoding the output at different depths of a pre-trained deep convolutional network using a Random-ized Autoencoder (RAE). The RAE is trained locally to each image using a closed-form solution, and its decoder weights are used to compose a 1-dimensional texture representation that is fed into a linear SVM. This means that no fine-tuning or backpropagation is needed for the backbone. We explore RADAM on several texture benchmarks and achieve state-of-the-art results with different com putational budgets. Our results suggest that pre-trained backbones may not require additional fine-tuning for texture recog-nition if their learned representations are better encoded. & COPY; 2023 Elsevier Ltd. All rights reserved. (AU)

• FAPESP's process: 22/03668-1 - Analysis of the Dynamic Behavior of Complex Systems and Artificial Neural Networks in Computer Vision and Artificial Intelligence
• Grantee: Kallil Miguel Caparroz Zielinski
• Support Opportunities: Scholarships in Brazil - Doctorate
• FAPESP's process: 21/09163-6 - Network science for optimizing artificial neural networks on computer vision
• Grantee: Leonardo Felipe dos Santos Scabini
• Support Opportunities: Scholarships abroad - Research Internship - Doctorate
• FAPESP's process: 18/22214-6 - Towards a convergence of technologies: from sensing and biosensing to information visualization and machine learning for data analysis in clinical diagnosis
• Grantee: Osvaldo Novais de Oliveira Junior
• Support Opportunities: Research Projects - Thematic Grants
• FAPESP's process: 21/07289-2 - Learning Representations using artificial neural networks and complex networks with applications in sensors and biosensors
• Grantee: Lucas Correia Ribas
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• FAPESP's process: 21/08325-2 - An analysis of network automata as models for biological and natural processes
• Grantee: Odemir Martinez Bruno
• Support Opportunities: Regular Research Grants
• FAPESP's process: 19/07811-0 - Artificial neural networks and complex networks: an integrative study of topological properties and pattern recognition
• Grantee: Leonardo Felipe dos Santos Scabini
• Support Opportunities: Scholarships in Brazil - Doctorate