Product Catalog Cleansing

Automate product classification, attribute extraction, and catalog enrichment using state-of-the-art multimodal AI. Fine-tuned SOTA OSS models on FireworksAI.

Powered by
Select Model
Description Style

Upload a product image or select from table below

Classification Results

📚 Example Products (Click a row to load image)

Click on any of the rows below to load the product image for analysis

Key Findings:

  • Qwen2.5-VL-72B-SFT achieves >95% accuracy on masterCategory
  • Fine-tuned model shows 18% improvement on subCategory vs base model
  • All models maintain >90% precision and recall on gender classification

🚀 Want to Fine-Tune Your Own Model? 🚀

Check out our comprehensive fine-tuning cookbook to learn how we built this model! The notebook covers:

  • Dataset preparation and exploration
  • Data preprocessing for VLM fine-tuning
  • Fine-tuning with Fireworks AI
  • Model evaluation and deployment

📓 Open Fine-Tuning Notebook

Perfect for adapting this approach to your own product catalog or domain-specific use case.