PyTorch Implementation of
"Rich Teacher Features for Efficient Single-Image Haze Removal"
This repository provides an efficient and lightweight framework for single-image dehazing, leveraging heterogeneous knowledge distillation from a lightweight super-resolution teacher model to guide a compact student dehazing network.
The method is simple yet effective, designed specifically for on-the-edge deployment where computational resources are limited.
Rich Teacher Features for Efficient Single-Image Haze Removal
Abstract:
Single-image haze removal is a long-standing hurdle for computer vision applications.
Several works have focused on transferring advances from image classification, detection, and segmentation to the niche of image dehazing, primarily focusing on contrastive learning and knowledge distillation. However, these approaches prove computationally expensive, raising concern regarding their applicability to on-the-edge use-cases. This work introduces a simple, lightweight, and efficient framework for single-image haze removal, exploiting rich "dark knowledge" information from a lightweight pre-trained super-resolution model via the notion of heterogeneous knowledge distillation. We designed a Feature Affinity Module to maximize the flow of rich feature semantics from the super-resolution teacher to the student dehazing network. In order to evaluate the efficacy of our proposed framework, its performance as a plug-and-play setup to a baseline model is examined. Our experiments are carried out on the RESIDE-Standard dataset to demonstrate the robustness of our framework to both synthetic and real-world domains. The extensive qualitative and quantitative results provided establish the effectiveness of the framework, achieving gains of up to 15% (PSNR) while reducing the model size by approximately 20×.