Semantic Furniture Finder is an AI-powered system designed to help customers find available furniture products by submitting an image or screenshot through a chatbot interface.
Instead of relying on visual image similarity, this project uses semantic understanding of furniture images to identify similar items — even if colors, models, or backgrounds differ.

ABC Furniture Company wants to implement a chatbot system on their website that enables customers to:

• Upload a screenshot or photo of furniture (e.g., a sofa).
• Receive suggestions of similar available furniture from the company’s catalog, along with product details.

Traditional image-matching approaches are often unreliable because:

• Users may upload images containing different models, colors, or backgrounds.
• Extra, irrelevant objects in photos can confuse visual models.
• Exact pixel-level matching isn’t effective for semantic similarity.

To overcome these challenges, the system uses semantic similarity instead of visual similarity.
By converting images into semantic vectors and comparing their meanings (not just their visuals), we achieve more accurate furniture recommendations.

• Front End:
  The user uploads a furniture image or screenshot via the chatbot interface.

• Semantic Furniture Finder (Django):
  Backend service managing all communication between the frontend, databases, and AI models.

• Fashion Descriptor:
  Extracts semantic descriptions from the user-submitted image.

• OpenAI API:
  Generates text-based semantic embeddings from the image input.

• Weaviate DB:
  Vector database used to store embeddings and perform semantic vector matching to find similar items.

• PostgreSQL DB:
  Stores product metadata and details for each item.

• Chatbot:
  Interacts with the user, providing product recommendations and descriptions based on semantic similarity.

1. User uploads an image of furniture to the chatbot.
2. Fashion Descriptor extracts a semantic description from the image.
3. OpenAI API generates embeddings (vector representation) of that description.
4. Weaviate DB searches for semantically similar furniture items in the company’s catalog.
5. PostgreSQL DB provides product details for the matched items.
6. Chatbot sends the top-matched results to the user with product information.

• Works even when user images have different angles, lighting, or colors.
• Provides meaningful matches through semantic understanding.
• Easily scalable with new furniture data.
• Integrates seamlessly with web-based chatbot systems.

• Add voice-based queries to improve user interaction.
• Integrate multimodal embeddings (image + text) for better semantic precision.
• Implement real-time product availability updates.

The Semantic Furniture Finder provides a smarter, AI-driven way for furniture retailers to help customers discover products that truly match their intent — not just their visuals.

Author: Deshitha Hansajith Senarath
Date: November 2025
Version: 1.0