README.md

metadata

sdk: docker

from pathlib import Path

readme = """# YOLO and Qwen2 VL Microplastic Detection

A computer vision web application for detecting microplastic and waste related objects from images, video frames, and browser based inputs. The project combines a custom YOLO model for object detection with a Qwen2 VL vision language model for risk analysis and recommendations.

Project overview

This project uses two AI stages.

1. YOLO detection

The YOLO model detects visible objects from the uploaded image or video frame. It returns object class, confidence score, bounding box, total count, and an annotated image.

2. VLM analysis

The Qwen2 VL model receives the original image and YOLO detection results. It then generates a risk level, explanation, and practical recommendations for microplastic contamination analysis.

Main features

1. Image based detection.
2. Video frame based detection.
3. YOLO object detection using a custom trained model.
4. Qwen2 VL based image reasoning.
5. Risk level prediction for environmental contamination.
6. Detection count and object list.
7. Flask based web interface.
8. Docker ready Hugging Face Spaces deployment.

Technology stack

1. Python
2. Flask
3. Ultralytics YOLO
4. OpenCV
5. PyTorch
6. Transformers
7. Qwen2 VL
8. Docker
9. Hugging Face Spaces

Project structure

CV Project Space
    README.md
    Dockerfile
    .dockerignore
    CV-Project
        app.py
        vlm_analyzer.py
        requirements.txt
        best.pt
        templates
            index.html
        preprocess.py
        model.py
        merge.py
        Inference.py
        varification.py