Summary of Dap-led: Learning Degradation-aware Priors with Clip For Joint Low-light Enhancement and Deblurring, by Ling Wang et al.
DAP-LED: Learning Degradation-Aware Priors with CLIP for Joint Low-light Enhancement and Deblurring
by Ling Wang, Chen Wu, Lin Wang
First submitted to arxiv on: 20 Sep 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary A novel transformer-based joint learning framework, DAP-LED, is proposed to comprehensively perceive diverse degradation levels at night for autonomous vehicles and robots. By leveraging Contrastive Language-Image Pretraining (CLIP) to adaptively learn the degradation levels from images at night, the framework jointly achieves low-light enhancement and deblurring. This enables learning rich semantic information and visual representation for optimization of downstream tasks such as depth estimation, segmentation, and detection in the dark. The approach introduces a CLIP-guided cross-fusion module to obtain multi-scale patch-wise degradation heatmaps from image embeddings, which are then fused via designed CLIP-enhanced transformer blocks to retain useful degradation information. Experimental results demonstrate state-of-the-art performance in the dark, with effective results for three downstream tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary DAP-LED is a new way to make cameras on cars and robots see better at night. Right now, these cameras struggle because of low light and blur from long exposure times. Other methods try to fix this by combining special models that enhance low-light images and remove blur. But these methods often don’t work well or add weird effects to the pictures. The researchers found that a different type of model, called CLIP, can actually learn how to see better in low light conditions. They created a new framework called DAP-LED that combines CLIP with other models to enhance images and remove blur. This helps cars and robots do tasks like measuring distance, identifying objects, and detecting motion more accurately at night. |
Keywords
» Artificial intelligence » Depth estimation » Optimization » Pretraining » Transformer
