VLM

• Brand Tagging with VLMs 2025-11-15

  Fast logo retrieval with SigLIP-2 embeddings, then strict JSON verification with LLaVA-OneVision-1.5. Uses a single Creative-Commons video as the running example.

• ClipTagger-12B VLM: Frame Captioning Tutorial 2025-11-01

  Step-by-step GPU-accelerated inference for ClipTagger-12b, including prompts, environment setup, and a BF16 PyTorch quickstart.