| --- |
| license: mit |
| tags: |
| - image-classification |
| - resnet50 |
| - human-nature-interaction |
| - urban-greenspace |
| library_name: pytorch |
| --- |
| |
| # HNIVision ResNet-50 (7-class scene classifier) |
|
|
| Fine-tuned ResNet-50 used by the [HNIVision](https://github.com/LabMingzeChen/HNIVision) |
| toolkit to classify social-media images into 7 Human-Nature Interaction |
| scene categories. |
|
|
| ## Classes |
|
|
| 1. Garden Vegetation |
| 2. Waterscape and Built Settings |
| 3. Pure Waterscape |
| 4. Built Environment |
| 5. Natural Landscape |
| 6. Mixed Urban Greenspace |
| 7. Park Activities |
|
|
| (Edit this list on the Hub web UI if your training classes differ.) |
|
|
| ## Usage |
|
|
| \`\`\`python |
| from hnivision.methods import Classification |
| clf = Classification() |
| result = clf.extract("image.jpg") |
| \`\`\` |
|
|
| Weights auto-download from this repo on first use if not in local cache. |
|
|
| ## Citation |
|
|
| Chen, M. (2026). HNIVision: Toolkit and benchmark for human-nature |
| interaction extraction from social-media imagery. |
|
|
