Leveraging ShuffleNet and LLaVA-Phi for State-of-the-Art Image Deblurring and Description for Mobile Devices

The growing demand for real-time image processing on edge devices calls for novel approaches that balance computational efficiency with high performance. This paper introduces an integrated solution combining ShuffleNet, a lightweight convolutional neural network, with LLaVA-Phi model for efficient image deblurring and descriptive analysis on mobile devices. ShuffleNet’s structural efficiency, characterized by channel shuffling and depth-wise convolutions, is exploited to deblur images swiftly, while LLaVA-Phi interprets the imagery to generate concise natural language descriptions. Our unified approach significantly enhances both the visual clarity of images and the accuracy of their associated descriptions with minimal computational overhead. Experimental results reveal substantial improvements over existing methods, confirming the efficacy of our approach for enhanced real-time image processing in computationally limited environments.