Amazon ML Challenge 2024

We participated in Amazon ML Challenge 2024 with a solution for extracting product attributes from images.

Team Q*

• Aman Prakash (Lead), NIAMT, Ranchi
• Sagnik Pramanik, Heritage Institute of Technology, Kolkata
• Ankit Rai, NIAMT, Ranchi
• Abhinav Sinha, BIT Mesra, Ranchi

Problem: Product Attribute Extraction from Images

Our approach uses the Moondream Vision Language Model (VLM), which processes images from the test.csv file to extract specific attributes like weight, dimensions, and more.

Key Highlights:

• Moondream VLM (1.6B parameters) was used for lightweight image-to-text processing.
• Extracted key product attributes using targeted prompts.
• Output cleaning and standardization were done using regex for consistency.

For more details, refer to the main script: main_team_qstart_amazonml.ipynb.