| title: HNIVision Demo | |
| emoji: 🌳 | |
| colorFrom: green | |
| colorTo: blue | |
| sdk: gradio | |
| sdk_version: 5.49.1 | |
| python_version: "3.11" | |
| app_file: app.py | |
| pinned: false | |
| license: mit | |
| short_description: Five vision methods for Human-Nature Interaction extraction | |
| # HNIVision Demo | |
| Interactive demo for the [HNIVision toolkit](https://github.com/LabMingzeChen/HNIVision). | |
| Runs **5 computer vision methods** on social-media imagery and merges them into a unified **4-layer Human-Nature Interaction schema**: | |
| - Image labeling (Google Vision) | |
| - Image classification (ResNet-50 fine-tuned on 7 HNI scenarios) | |
| - Object detection (YOLO26-s) | |
| - Image segmentation (SAM2 + SegFormer hybrid on ADE20K) | |
| - Vision-LLM (GPT-4.1-mini) | |