README.md

metadata

title: Dress Detection and Classification Pipeline
emoji: 👗
colorFrom: pink
colorTo: purple
sdk: gradio
sdk_version: 4.44.0
app_file: app.py
pinned: false
license: mit

Dress Detection and Classification Pipeline

This application performs end-to-end dress analysis using computer vision and deep learning:

Features

1. Human Detection: Detects humans in uploaded images using Faster R-CNN
2. Dress Segmentation: Segments the dress area using a specialized U-Net model
3. Classification: Classifies the dress type using a ConvNeXt model
4. Grad-CAM: Shows attention areas for the classification decision with interpretability

How to Use

1. Upload an image containing a person wearing a dress
2. The pipeline will automatically:
   • Detect the person in the image
   • Segment the dress area
   • Classify the dress type
   • Show what parts of the dress the model focused on for classification

Technical Details

• Detection Model: Faster R-CNN with ResNet-50 backbone
• Segmentation Model: U-Net with ResNet-34 encoder
• Classification Model: ConvNeXt v2 Base
• Interpretability: Grad-CAM for attention visualization

Note

This is an academic project demonstrating the integration of multiple computer vision models in a single pipeline. The models may require fine-tuning for optimal performance on specific datasets.