Downloads · 30 days
0
aapdasetu-damage-dataset
Divyanshu-Kumar19/aapdasetu-damage-dataset
aapdasetu-damage-dataset is a machine learning model from Divyanshu-Kumar19. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as mit.
Ground-level photos of disaster-affected buildings, labelled by damage severity. Used to train the ResNet50 damage classifier Divyanshu-Kumar19/aapdasetu-damage-assessment (98.36% test accuracy with 5-crop TTA).
Access
Public
Updated Aug 8, 2026
Repo size
62 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.jpg136 MB · 100%
At a glance
- License
- mit
- Access
- Public
- Created
- Aug 8, 2026
- Updated
- Aug 8, 2026
- SHA
- 635a9d11
Model card
From the Hugging Face model README
AapdaSetu Building Damage Dataset
Ground-level photos of disaster-affected buildings, labelled by damage severity. Used to train the ResNet50 damage classifier Divyanshu-Kumar19/aapdasetu-damage-assessment (98.36% test accuracy with 5-crop TTA).
Part of the AapdaSetu project (Smart India Hackathon).
Classes
| Class | Count | Description |
|---|---|---|
MINOR | 846 | Hairline cracks, broken windows, superficial damage — structure liveable |
MAJOR | 525 | Partial wall/roof collapse, severe damage — main structure still exists |
DESTROYED | 1,029 | Complete collapse, rubble only |
Layout
raw/
├── MINOR/ (846 images)
├── MAJOR/ (525 images)
└── DESTROYED/ (1,029 images)
splits.json records the class-stratified split actually used for training:
75% train / 15% val / 10% test per class, with a perceptual-hash leak check
verifying no image (or near-duplicate) appears in more than one split.
Usage
from torchvision.datasets import ImageFolder
dataset = ImageFolder("raw") # class_to_idx is alphabetical:
# DESTROYED=0, MAJOR=1, MINOR=2
Or download with the HF CLI:
hf download Divyanshu-Kumar19/aapdasetu-damage-dataset --local-dir ./dataset
Files
| File | Description |
|---|---|
raw/{MINOR,MAJOR,DESTROYED}/*.jpg | Labelled source images |
splits.json | Train/val/test split counts per class |
DATASET_GUIDE.md | Guide on collecting and preparing your own photos |
Related
- Model + checkpoint: Divyanshu-Kumar19/aapdasetu-damage-assessment
- Service code:
ai-service/in the SIH repo (divyanshubranch)
License
MIT
- License
- mit
- Created
- Aug 8, 2026
- Updated
- Aug 8, 2026