SnapGrade: A Classical CV Photo Intelligence Pipeline That Runs at 3.3 img/s on 8 GB RAM
Why classical CV — not a VLM — is the right tool for photo culling. Six metrics, a 40-line decision layer, and 3.3 img/s on 8 GB of RAM.
// senior data scientist & ai/ml engineer
Technical blog by Bibhash Mitra covering AI, Machine Learning, Deep Learning, and Data Science projects.
03 / Posts
Deep dives into AI/ML projects, tutorials, and insights
Why classical CV — not a VLM — is the right tool for photo culling. Six metrics, a 40-line decision layer, and 3.3 img/s on 8 GB of RAM.
A free local web app that does smart photo culling — blurry shots, closed eyes, burst picks — without uploading a single file to the cloud.
Stop searching for code. Start repolecting about it 😉. Repolect indexes your codebase into a hierarchical semantic tree and knowledge graph, then navigates it with LLM reasoning...
Learn how to host IBM's Docling-Serve on Hugging Face Spaces for free. A complete walkthrough of Docker configuration, rate limiting, and API usage for document AI.
A comprehensive playbook for upgrading from vanilla RAG to state-of-the-art document QA -- covering contextual embeddings, hybrid retrieval, reranking, GraphRAG, and self-correc...
04 / Browse
Filter posts by topic area