Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
NVIDIA Cosmos: Lab to Field
Fix data scarcity and domain shift by translating rare moth lab images to realistic field data using NVIDIA Cosmos-Transfer2.5 and validating results in FiftyOne.
A hands-on demo showing how to solve real-world data scarcity and domain shift using NVIDIA Cosmos-Transfer2.5 and FiftyOne. We begin by extracting rare moth samples from the BioTrove biodiversity dataset using BioCLIP embeddings, converting them into controllable video inputs, and generating edge-based control signals. NVIDIA Cosmos-Transfer2.5 then performs world-model–driven domain translation, transforming sterile lab images into realistic field environments. The outputs are validated and analyzed in FiftyOne using grouped datasets, embeddings, and semantic similarity search. This generative workflow yields 20–40% improvements on downstream classifiers by expanding rare classes and correcting dataset bias.
This demo directly addresses three core challenges in real-world computer vision: (1) severe class imbalance and ecological domain shift, (2) absence of realistic field imagery for rare species, and (3) scalable validation of generative outputs. The session walks through the pipeline end-to-end: selecting scarce samples with BioCLIP, preparing video and edge-control inputs, running Cosmos-Transfer2.5 for domain transfer, and validating the generated outputs in FiftyOne using embeddings, grouped datasets, and semantic search.
Related projects
Meet DARIA, the cutest little AI that's about to make grading disappear!
Raleigh
Demo the DARIA prototype, detailing the multi-modal data pipeline and specialized model fine-tuning for assessing complex project submissions…
Glimpse of the future: How hyperspeed inference allows for new class of ai driven apps
Raleigh
Explore how 1000 tokens/s inference enables new AI app designs, like on-demand OS tools, discussing agentic AI and…
Think Outside the "Blocks": You Don’t Need a PhD to Solve Novel Problems in AI
Raleigh
Learn how a non-expert used LLMs to build a novel tool for generating blocky 3D models for games,…
Fitness Tracking Using Computer Vision
Raleigh
Learn how to build a non-contact computer vision algorithm using mobile phones to track and analyze deadlift movements,…
Stop Cloud Waste: AI Checks Costs Before Deploy in AWS
Raleigh
AI reviews cloud infrastructure plans before deployment, identifying cost overruns and policy violations with personalized suggestions based on…
From Local Prototyping to Distributed Clusters: An Open Source Platform for ML Research Teams
Toronto
See a demo scaling ML training from a local notebook to a GPU cluster, covering checkpoint recovery, hyperparameter…