# ADAM-Dehaze: Adaptive Density-Aware Multi-Stage Dehazing for Improved Object Detection in Foggy Conditions

arXiv: https://arxiv.org/abs/2506.15837 · HTML: https://mlsanity.com/paper/2506.15837/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 22 of its 38 references that are in the map.
- First posted on arXiv 16 Jun 2025.

## Builds on (most cited first, 22 of 22)

- [Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks](https://mlsanity.com/paper/1506.01497/) (2015)
- [The Cityscapes Dataset for Semantic Urban Scene Understanding](https://mlsanity.com/paper/1604.01685/) (2016)
- [SSD: Single Shot MultiBox Detector](https://mlsanity.com/paper/1512.02325/) (2015)
- [YOLOv3: An Incremental Improvement](https://mlsanity.com/paper/1804.02767/) (2018)
- [Benchmarking Single Image Dehazing and Beyond](https://mlsanity.com/paper/1712.04143/) (2017)
- [NIMA: Neural Image Assessment](https://mlsanity.com/paper/1709.05424/) (2017)
- [Benchmarking Robustness in Object Detection: Autonomous Driving when Winter is Coming](https://mlsanity.com/paper/1907.07484/) (2019)
- [DehazeNet: An End-to-End System for Single Image Haze Removal](https://mlsanity.com/paper/1601.07661/) (2016)
- [FFA-Net: Feature Fusion Attention Network for Single Image Dehazing](https://mlsanity.com/paper/1911.07559/) (2019)
- [Multi-Scale Boosted Dehazing Network with Dense Feature Fusion](https://mlsanity.com/paper/2004.13388/) (2020)
- [GridDehazeNet: Attention-Based Multi-Scale Network for Image Dehazing](https://mlsanity.com/paper/1908.03245/) (2019)
- [Gated Fusion Network for Single Image Dehazing](https://mlsanity.com/paper/1804.00213/) (2018)
- [Deep Generalized Unfolding Networks for Image Restoration](https://mlsanity.com/paper/2204.13348/) (2022)
- [Domain Adaptation for Image Dehazing](https://mlsanity.com/paper/2005.04668/) (2020)
- [Synscapes: A Photorealistic Synthetic Dataset for Street Scene Parsing](https://mlsanity.com/paper/1810.08705/) (2018)
- [Model Adaptation with Synthetic and Real Data for Semantic Dense Foggy Scene Understanding](https://mlsanity.com/paper/1808.01265/) (2018)
- [RIDCP: Revitalizing Real Image Dehazing via High-Quality Codebook Priors](https://mlsanity.com/paper/2304.03994/) (2023)
- [End-to-End United Video Dehazing and Detection](https://mlsanity.com/paper/1709.03919/) (2017)
- [Semantic Understanding of Foggy Scenes with Purely Synthetic Data](https://mlsanity.com/paper/1910.03997/) (2019)
- [Haze Visibility Enhancement: A Survey and Quantitative Benchmarking](https://mlsanity.com/paper/1607.06235/) (2016)
- [FLARE up your data: Diffusion-based Augmentation Method in Astronomical Imaging](https://mlsanity.com/paper/2405.13267/) (2024)
- [Optimizing Wireless Networks with Deep Unfolding: Comparative Study on Two Deep Unfolding Mechanisms](https://mlsanity.com/paper/2403.18930/) (2024)

## Built on by (most cited first, 1 of 1)

- [C2FXNet: Coarse-to-Fine Scene Expert for Unified Object Detection across Adverse Weather](https://mlsanity.com/paper/2609.25693/) (2026)

## Abstract (arXiv)

Adverse weather conditions, particularly fog, pose a significant challenge to autonomous vehicles, surveillance systems, and other safety-critical applications by severely degrading visual information. We introduce ADAM-Dehaze, an adaptive, density-aware dehazing framework that jointly optimizes image restoration and object detection under varying fog intensities. A lightweight Haze Density Estimation Network (HDEN) classifies each input as light, medium, or heavy fog. Based on this score, the system dynamically routes the image through one of three CORUN branches: Light, Medium, or Complex, each tailored to its haze regime. A novel adaptive loss balances physical-model coherence and perceptual fidelity, ensuring both accurate defogging and preservation of fine details. On Cityscapes and the real-world RTTS benchmark, ADAM-Dehaze improves PSNR by up to 2.1 dB, reduces FADE by 30 percent, and increases object detection mAP by up to 13 points, while cutting inference time by 20 percent. These results highlight the importance of intensity-specific processing and seamless integration with downstream vision tasks. Code available at: https://github.com/talha-alam/ADAM-Dehaze.