ehsanulhaque92/AuraScanAI

1---2title: AuraScanAI API3emoji: 🚗4colorFrom: blue5colorTo: purple6sdk: docker7app_port: 50008---9 10# AuraScanAI - Vehicle Damage Assessment API11 12This repository contains the complete backend service for the AuraScanAI project, a sophisticated AI-powered system for analyzing vehicle damage from images. The API is built with Flask and serves a custom-trained, multi-task Vision Transformer (ViT) model capable of identifying damage areas and assessing their severity.13 14The live API is deployed as a Docker container on Hugging Face Spaces.15 16---17 18## 🚀 Key Features19 20*   **AI-Powered Analysis:** Leverages a state-of-the-art Vision Transformer (ViT) model fine-tuned on over 15,000 images of vehicle damage.21*   **Multi-Task Learning:** The model simultaneously predicts:22    1.  The location of the primary damage area (bounding box).23    2.  The overall severity of the damage (`minor`, `moderate`, `severe`).24*   **Business Logic Engine:** Includes a post-processing layer to translate AI outputs into actionable business insights, including a realistic estimated repair cost range.25*   **Scalable Architecture:** Built with a professional, singleton pattern to ensure the large AI model is loaded only once, providing fast and efficient inference.26*   **Containerized & Deployable:** Fully containerized with Docker and configured for seamless deployment on cloud platforms like Hugging Face Spaces.27 28---29 30## 🛠️ Technology Stack31 32*   **AI Framework:** PyTorch33*   **Vision Model Library:** `timm` (PyTorch Image Models)34*   **API Framework:** Flask35*   **WSGI Server:** Gunicorn36*   **Containerization:** Docker37*   **Cloud Deployment:** Hugging Face Spaces38 39---40 41## ⚙️ API Endpoints42 43The server provides two main endpoints:44 45### 1. Health Check46 47A simple endpoint to verify that the server is running and responsive.48 49*   **Endpoint:** `/ping`50*   **Method:** `GET`51*   **Success Response (200):**52    ```json53    {54      "message": "Server is alive!",55      "status": "ok"56    }57    ```58 59### 2. Damage Analysis60 61The core endpoint for analyzing an image.62 63*   **Endpoint:** `/analyze`64*   **Method:** `POST`65*   **Request Body:** `multipart/form-data` with a single field:66    *   `file`: The vehicle image file (`.jpg`, `.png`, etc.).67*   **Success Response (200):** A detailed JSON object containing the full analysis.68    ```json69    {70      "success": true,71      "totalDamages": 1,72      "overallSeverity": "severe",73      "confidence": "0.66",74      "costRange": {75        "min": 800,76        "max": 250077      },78      "damages": [79        {80          "id": "dmg-1",81          "type": "Primary Damage Area",82          "location": "Detected by AI",83          "severity": "severe",84          "estimatedCost": { "min": 800, "max": 2500 },85          "coordinates": [86            [ 423.06, 364.49, 1229.91, 859.14 ]87          ]88        }89      ]90    }91    ```92*   **Error Response (4xx/5xx):**93    ```json94    {95      "success": false,96      "error": "Descriptive error message."97    }98    ```99 100---101 102## 📜 MVP Approach & Future Roadmap103 104This project serves as a powerful Proof of Concept (MVP), demonstrating a complete end-to-end pipeline for AI-powered vehicle damage assessment.105 106**Current Capability (MVP):**107The current AI model is an **Image Assessment Model**, designed to identify the single most prominent damage area in an image. It provides a holistic analysis, including an overall severity classification, estimated repair cost, and a bounding box for the primary damage region. This successfully proves the core technology is viable.108 109**Future Roadmap:**110The next phase of this project will involve evolving the AI core into a full **Multi-Object Detector** (e.g., using a DETR or YOLO architecture). This will enable the system to:111-   Identify and draw bounding boxes for multiple, distinct damages in a single image.112-   Provide a detailed breakdown and cost estimate for each individual damage in the Damage Ledger.