Advances in Image Restoration and Continual Learning

The field of image restoration and continual learning is rapidly advancing, with a focus on developing innovative methods to improve the accuracy and efficiency of these tasks. Recent research has explored the use of reinforcement learning, adaptive total variation regularization, and coreset selection to enhance image restoration and continual learning capabilities. Notably, the development of frameworks such as FoodRL, IMDNet, and TimeSearch-R has demonstrated significant potential for social impact and adaptive ensemble learning. Additionally, the introduction of methods like PDAC and CADIC has improved the efficiency and effectiveness of coreset selection and continual anomaly detection. Overall, the field is moving towards more sophisticated and adaptive approaches to image restoration and continual learning, with a emphasis on real-world applications and social impact.

Noteworthy papers include: FoodRL, which proposes a novel reinforcement learning framework for in-kind food donation forecasting, demonstrating significant potential for social impact. IMDNet, which introduces an adaptive multi-degradation image restoration network, showing superior performance on multi-degradation restoration tasks. TimeSearch-R, which reformulates temporal search as interleaved text-video thinking, achieving state-of-the-art performance on temporal search benchmarks.

Sources

FoodRL: A Reinforcement Learning Ensembling Framework For In-Kind Food Donation Forecasting

Learning to Restore Multi-Degraded Images via Ingredient Decoupling and Task-Aware Path Adaptation

Real-World Adverse Weather Image Restoration via Dual-Level Reinforcement Learning with High-Quality Cold Start

PreResQ-R1: Towards Fine-Grained Rank-and-Score Reinforcement Learning for Visual Quality Assessment via Preference-Response Disentangled Policy Optimization

Sharing the Learned Knowledge-base to Estimate Convolutional Filter Parameters for Continual Image Restoration

TimeSearch-R: Adaptive Temporal Search for Long-Form Video Understanding via Self-Verification Reinforcement Learning

The Online Patch Redundancy Eliminator (OPRE): A novel approach to online agnostic continual learning using dataset compression

CADIC: Continual Anomaly Detection Based on Incremental Coreset

An ICTM-RMSAV Framework for Bias-Field Aware Image Segmentation under Poisson and Multiplicative Noise

Compact Memory for Continual Logistic Regression

A Semi-Convergent Stage-Wise Framework with Provable Global Convergence for Adaptive Total Variation Regularization

PDAC: Efficient Coreset Selection for Continual Learning via Probability Density Awareness

Built with on top of