Pooh1101/All-Around-Assistance-AAA-Dataset

1

stars

19

commits

Python

primary language

Nov 5, 2025

updated

README

All Around Assistance (AAA) Dataset

About the AAA Dataset

AAA is an open-source dataset for vision-and-language research and accessibility. It provides four Angles of View (AoV) to enable controlled evaluation across camera geometries: 70°, 110°, 180° (fisheye), and 360° (equirectangular).

To support assistive applications for visually impaired users, AAA also includes a targeted Q&A benchmark organized into four categories:

  • Object — identify, locate, or describe items.
  • Safety — detect obstacles, or risky conditions.
  • Navigation — directions, and landmarks for wayfinding.
  • Surrounding — scene context, and overall environment.

Together, these components enable fair comparison of models across AoVs while evaluating practical, assistive questions that matter in real-world settings.

We do not redistribute GND. Please download from the official site: GND dataset.

Quick Start

git clone https://github.com/Pooh1101/All-Around-Assistance-AAA-Dataset.git
cd All-Around-Assistance-AAA-Dataset

Download GND Dataset

Ensure the structure is as follows:

frames_folder_root/
├── AU/
│   ├── S01/
│   │   ├── frame0000
│   │   ├── frame0001
│   │   └── ...
│   ├── S02/
│   │   └── ...
│   └── ...
│
└── CUA/
    └── ... 

# build views (70 / 110 / 180 / 360)
python make_views_min.py </path/to/frames_folder_root> </path/to/output_folder>

#run Mask2Former
python mask2former_build_masks.py </path/to/output_folder>/frames --model facebook/mask2former-swin-large-coco-panoptic

Contributors

Pooh1101

19 commits

Pooh1101/All-Around-Assistance-AAA-Dataset

1

stars

19

commits

Python

primary language

Nov 5, 2025

updated

README

All Around Assistance (AAA) Dataset

About the AAA Dataset

AAA is an open-source dataset for vision-and-language research and accessibility. It provides four Angles of View (AoV) to enable controlled evaluation across camera geometries: 70°, 110°, 180° (fisheye), and 360° (equirectangular).

To support assistive applications for visually impaired users, AAA also includes a targeted Q&A benchmark organized into four categories:

  • Object — identify, locate, or describe items.
  • Safety — detect obstacles, or risky conditions.
  • Navigation — directions, and landmarks for wayfinding.
  • Surrounding — scene context, and overall environment.

Together, these components enable fair comparison of models across AoVs while evaluating practical, assistive questions that matter in real-world settings.

We do not redistribute GND. Please download from the official site: GND dataset.

Quick Start

git clone https://github.com/Pooh1101/All-Around-Assistance-AAA-Dataset.git
cd All-Around-Assistance-AAA-Dataset

Download GND Dataset

Ensure the structure is as follows:

frames_folder_root/
├── AU/
│   ├── S01/
│   │   ├── frame0000
│   │   ├── frame0001
│   │   └── ...
│   ├── S02/
│   │   └── ...
│   └── ...
│
└── CUA/
    └── ... 

# build views (70 / 110 / 180 / 360)
python make_views_min.py </path/to/frames_folder_root> </path/to/output_folder>

#run Mask2Former
python mask2former_build_masks.py </path/to/output_folder>/frames --model facebook/mask2former-swin-large-coco-panoptic

Contributors

Pooh1101

19 commits

Languages

Python

100.0%