---
title: Inference Studio | AI Vision Explorer
emoji: 🚀
colorFrom: blue
colorTo: purple
sdk: docker
app_port: 7860
pinned: false
---

# 🚀 Inference Studio | AI Vision Explorer

A premium, web-based interface for deploying and testing YOLO vision models with advanced controls for ROI (Region of Interest) filtering and confidence range analysis.

![UI Preview](static/ui_preview.png)

## ✨ Features

-   **Interactive ROI Drawing**: Draw detection zones directly on a preview image or video frame.
-   **Confidence Range Filtering**: Test model behavior by specifying both Min and Max confidence thresholds (e.g., visualize only low-confidence detections).
-   **Video Studio**: Full support for video inference with automatic frame extraction for ROI setup and H.264 transcoding for web playback.
-   **Glassmorphism UI**: Modern, dark-themed interface built for a premium developer experience.
-   **Model Management**: Easily swap `.pt` models on the fly.

## 🛠️ Setup

### Prerequisites
-   **Python 3.8+**
-   **FFmpeg**: Required for video processing.
    ```bash
    # Ubuntu/Debian
    sudo apt update && sudo apt install ffmpeg
    ```

### Installation
1.  Clone the repository:
    ```bash
    git clone https://github.com/your-repo/inference-studio.git
    cd inference-studio
    ```
2.  Install dependencies:
    ```bash
    pip install -r requirements.txt
    ```

## 🚀 Running the Studio

Start the server using the provided shell script:
```bash
chmod +x start.sh
./start.sh
```
The studio will be available at `http://localhost:8000`.

## 📂 Project Structure

-   `main.py`: FastAPI backend handling inference and video tasks.
-   `static/`: Frontend assets (styles, interactive JS).
-   `templates/`: HTML templates.
-   `uploads/`: Directory structure for models, videos, and results (ignored by Git).

## 📝 License
MIT