ClipScrape

A Streamlit-based web app for scraping images from any webpage, filtering out advertisements using a ResNet model, and enabling semantic image and text-based search using OpenAI's CLIP and FAISS.

Clip Scraper

Visionary Scraper is an AI-powered web application that allows you to:

• Scrape images from any given URL.
• Automatically filter out advertisement images using a ResNet classifier.
• Generate image embeddings using OpenAI’s CLIP model.
• Perform semantic search via image or text queries using FAISS.
• All wrapped inside an easy-to-use Streamlit interface.

Features

• Image Scraping: Fetch all images from a public webpage.
• Ad Filtering: Filter out advertisement images using a pre-trained ResNet-18 model.
• CLIP Embeddings: Generate embeddings using OpenAI’s CLIP model (clip-vit-base-patch32).
• Semantic Search:
  • Image-to-Image Search: Find visually or semantically similar images.
  • Text-to-Image Search: Retrieve images that best match a given text prompt.
• Fast Retrieval: Use FAISS for efficient nearest-neighbor search on high-dimensional embeddings.

Technologies Used

Library	Purpose
Streamlit	Web app interface
BeautifulSoup, requests	Web scraping
torch, torchvision	Deep learning models
transformers	CLIP model from Hugging Face
PIL	Image loading and processing
faiss	Efficient similarity search
datasets	Batch encoding and loading

Installation

1. Clone the repository

   git clone https://github.com/yourusername/visionary-scraper.git
   cd visionary-scraper