MODEL_CARD.md

metadata

language:
  - en
license: mit
tags:
  - computer-vision
  - clip
  - deduplication

AI Visual Gallery

This project implements an image deduplication pipeline using visual features extracted by CLIP (Contrastive Language-Image Pre-training).

Methodology

1. Feature Extraction: Images are passed through a Vision Transformer (ViT).
2. Indexing: 512-dim feature vectors are cached.
3. Comparison: Cosine similarity is used to detect duplicates.

Hardware Requirements

• RAM: 2GB+ (for caching feature maps)