BAGHDAD-ADS: Benchmark Dataset for Baghdad Adverse Driving Scenes

BAGHDAD-ADS is a curated dataset of real urban driving scenes captured in Baghdad, Iraq. It is designed to support image enhancement and autonomous-driving perception research under challenging visual conditions, including night scenes, headlight glare, rain streaks, rain smear artifacts, and fog.

Dataset Samples

Sample visual conditions from BAGHDAD-ADS, including night, nighttime glare, rain streaks, rain smear artifacts, and fog scenes collected from Baghdad roads.

BAGHDAD-ADS dataset visual conditions including night, nighttime glare, rain streaks, rain smear artifacts, and fog

Prepared by:

Yousif N. Abbas

AI and Computer Vision Researcher specialized in image enhancement and autonomous driving systems.

Visit Profile

Dataset Information

Weather Condition Number of Samples
Rain Streaks 4,434
Nighttime Glare 667
Night 4,921
Fog 500
Rain Smear 11,521

How to Get Access?

PNG Format
v1.1 (Current) 3.74 GB Released May 2026
Contact me at me@yousif.app to request access.

Paired Dataset Samples

The BAGHDAD-ADS dataset provides unique paired driving sequences to facilitate advanced research in computer vision and image restoration. Each sample includes degraded input scenes captured under adverse environmental conditions such as heavy fog, torrential rain, and low-light night driving, matched perfectly with high-quality reference/ground-truth images. This pairing is essential for training and evaluating supervised deep learning models focused on dehazing, deraining, and low-light enhancement.

BAGHDAD-ADS paired input and reference samples
Sample visualization: The top row displays various degraded inputs under glare, rain, fog, low-light, while the bottom row shows the corresponding high-quality reference frames used as ground truth.

Citation

If you use this dataset in your research, please cite the following publication:

@article{abbas_uwunet_2026,
  title={Unified Wavelet U-Net with Bottleneck Classification for Adaptive Image Enhancement in Autonomous Driving Systems},
  author={Abbas, Yousif Neamah and Abdulmunim, Matheel Emaduldeen and Ali, Nada Hussain and Ahmad, Musheer and Abdel Mageed, Ismail},
  journal={International Journal of Intelligent Engineering and Systems},
  volume={19},
  number={6},
  pages={105--126},
  year={2026},
  doi={10.22266/ijies2026.0630.07},
  url={https://inass.org/wp-content/uploads/2026/03/2026063007-2.pdf}
}