Improving Qwen-VL Structured Output with Image Cropping

Explaining how I'm improving structured output results from Qwen-VL with image cropping in Sparrow.

 

 

Document Querying with Qwen2-VL-7B and JSON Output

In this video, I demonstrate how to perform document queries using Qwen2-VL-7B. By simplifying field names, we streamline the prompts, making them more efficient and reusable across different documents. This approach is similar to running SQL queries on a database, but tailored for language models like Qwen2-VL-7B, with results returned in JSON format.