Lightweight Deep Learning Model for Lung X-Ray Classification on Edge Devices.

Category: *Precision Health & Smart Medical

Exhibitor: NATIONAL TAIWAN NORMAL UNIVERSITY

Booth No: N615

Characteristic

This study proposes a lightweight deep learning model deployable on edge devices for multi-class classification of lung diseases from chest X-ray images. The model integrates decomposed convolution and symmetric separable convolution (SSC Block) to significantly reduce parameters and computational cost while maintaining high accuracy. Coordinate Attention is incorporated to enhance the recognition of lesion locations and shapes. Training data is balanced using CLAHE enhancement and MixUp/CutMix augmentation, enabling the model to classify 12 disease categories with an average accuracy of 91%. The model has been successfully deployed on devices like the Raspberry Pi, demonstrating real-time inference capability and strong potential for mobile healthcare and resource-limited settings.

Menu

Lightweight Deep Learning Model for Lung X-Ray Classification on Edge Devices.

Characteristic

Other Products

Products you may be interested in

Highest Rated Products