NVIDIA Speeds Up Bounding Box Detection 10x by Redefining VLM Grounding
NVIDIA has achieved a breakthrough in computer vision, speeding up bounding box detection by 10x by eliminating a traditionally mandatory step in VLM grounding.
Sources x.com
Marlin-2B: An ultra-compact Vision-Language Model for video information extraction
Marlin-2B is an open-source VLM with only 2 billion parameters that offers powerful video analysis capabilities, competing directly with larger models like Gemini-2.5-flash.
Sources x.com