A Hybrid Human-Machine System for Image-Based Multi-weather Detection
Document Type
Conference Article
Publication Title
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Abstract
Accurate determination of weather from images or video is of prime significance in applications such as autonomous vehicles drive or atmospheric pollution estimation. In contrast to image classification approaches generally adopted in the literature based on neural networks acting directly on the input, we propose a novel joint learning approach combining features extracted from the input image, sensitive to the human visual system (HVS) and those generated by the CNN model trained on benchmark datasets. The features work in a joint collaboration that is able to detect the presence of weather features during learning. The novel approach outperforms many state-of-the-art methods which use only CNN-based extracted features in order to classify the images. Experimental results with publicly available benchmark datasets establish the robustness and effectiveness of the proposed method in multi-weather classification.
First Page
315
Last Page
329
DOI
10.1007/978-3-031-25825-1_23
Publication Date
1-1-2023
Recommended Citation
Bhandari, Harsh; Chowdhury, Soumajit; and Palit, Sarbani, "A Hybrid Human-Machine System for Image-Based Multi-weather Detection" (2023). Conference Articles. 632.
https://digitalcommons.isical.ac.in/conf-articles/632