Roboflow
Day 34 after signup
Action recognition is now in Roboflow
Live today: video reasoning with NVIDIA Cosmos 3
Why it works
- Lead with one concrete product capability and explain what it enables before offering a deeper-read link.
- Group secondary items under clear labels so readers can scan a broad roundup and choose the topics they care about.
How it is built
- New capability introduction
- Livestream invitation
- Recorded talks roundup
- Community spotlights
Email copy
Catch every missed step on the line Launch Turn videos into insights Roboflow now supports action recognition, so a model can watch your video and return each action with its start and end time. Label time ranges, train Cosmos 3 Edge or V-JEPA 2.1, and deploy to the cloud or your own GPUs. Read the Blog Livestream Verify assembly steps and cycle times on video Join NVIDIA and Peter Robicheaux, Models Team Lead at Roboflow, live today at 9 a.m. PT (12 p.m. ET). Watch video reasoning with Cosmos 3 Edge in Roboflow, then learn to fine-tune it on your own videos to recognize actions and when they occur. Join the Livestream Video Watch talks from Tesla, Meta, NVIDIA, and more All 13 talks from the Visual Intelligence Summit are ready to watch. Speakers from BNSF, Driveline Baseball, and more share how they use vision across rail, logistics, manufacturing, and sports. Watch the Sessions Community spotlights Gemini 4 Carbon leads Playground evals Piotr Skalski finds Gemini 4 Carbon ranks first overall in Playground, with top scores in OCR and long text transcription. (X) Tell players apart across game clips Alex Bodner shares reid, Roboflow's new package that turns object crops into embeddings to match players across clips. (X) Roboflow project of the week Flag surface cracks and corrosion on wind turbines, grade weld quality, or log damage across a full inspection with this Defect Detection of Wind Turbine object detection API. View on Roboflow Universe → Top new releases RF-DETR 1.12.0: one prediction API, INT8 exports, and faster inference → GitHub Claude Fable 5.2: better chart reading, coding, and reasoning → Anthropic Research of the week Annotator-free object detection via foundation model collaboration → arXiv Are frontier VLM agents ready to be robot generalists? → arXiv © 2026 Roboflow, Inc. All rights reserved. 500 Locust St, Suite 104, Des Moines, IA 50309 Manage Preferences | Unsubscribe