Adaptive sparse coding improves robustness of visual recognition systems
Adaptive Convolutional Sparse Coding via Information Bottleneck for Robust Visual Signal Representation
Computer Vision and Pattern Recognition
Summary
Visual information needs to be stored in a way that keeps important details but removes unnecessary ones. The authors created a method that automatically adjusts how much information to keep when analyzing images, instead of using a fixed setting. They use a technique that balances storing enough information for the task but compressing to reduce noise or corruption. This method shows better performance on recognizing images, especially when the input is noisy or altered.
What this means in practice
- •For computer vision engineers: Enhance image recognition models to be more resistant to corrupted or noisy visual inputs by dynamically adjusting sparse coding parameters.
- •For mobile device developers: Improve robustness of onboard image processing under varying and noisy real-world conditions by adapting compression strength without retraining models.$Commercial implications: This enables more reliable image-based applications on mobile devices by automatically tuning compression for noise, improving user experience and device performance.
Authors
Meng'en Qin, Yinchen Liu, Mingxuan Cui, Youlu Xing
Abstract
Visual signals require compact yet sufficient representations for robust downstream prediction. Convolutional sparse coding (CSC) provides an explicit mechanism for suppressing redundant components while preserving signal content, but its sparsity coefficient is typically fixed and manually selected. We propose an adaptive convolutional sparse coding framework for robust visual signal representation. Specifically, we unfold the CSC optimization with the Fast Iterative Shrinkage-Thresholding Algorithm (FISTA) and treat the sparsity coefficient as a differentiable variable jointly learned with the network parameters. From the information bottleneck perspective, this coefficient controls the trade-off between information retention and compression: the sparsity term promotes compact representations, while the reconstruction term together with task loss preserves task-relevant signal content. We further introduce a label-free post-training strategy that adjusts the compression strength for corrupted inputs with the main network parameters fixed. Experiments on CIFAR and ImageNet demonstrate competitive clean-data recognition and greatly improved robustness under different input perturbations.