Upload README.md with huggingface_hub

README.md

@@ -1,10 +1,42 @@
 ---
-title: Demomule
-emoji: 🦀
-colorFrom: green
-colorTo: pink
+title: SAM3 Training Data Accelerator
+emoji: 🏷️
+colorFrom: blue
+colorTo: purple
 sdk: docker
-pinned: false
+app_port: 8501
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# SAM3 Training Data Accelerator
+
+Accelerate training data creation for object detection models using Facebook SAM3 promptable segmentation.
+
+## Workflow
+
+1. **Upload** an image
+2. **Detect** objects via text prompt (e.g. "individual servers mounted in a rack")
+3. **Label** each detection — accept or reject, assign a class label
+4. **Refine** using accepted boxes as prompts to find additional objects
+5. **Export** a JSON manifest for downstream fine-tuning
+
+## Deployment
+
+### HuggingFace Spaces
+
+Push this directory to an HF Space with **Docker** SDK and GPU runtime. Set `HF_TOKEN` as a Space secret (required for gated `facebook/sam3` model access).
+
+### Local / Docker
+
+```bash
+docker build -t demomule .
+docker run -p 8501:8501 -e HF_TOKEN=hf_... demomule
+```
+
+Then open http://localhost:8501.
+
+### Local (no Docker)
+
+```bash
+pip install -r requirements.txt
+HF_TOKEN=hf_... streamlit run app.py
+```