Cosmos Transfer 2.5 agtec single-view post-training recipe
NVIDIA + Aigen
Robot vision teams never have enough footage of unusual field situations. This is an open recipe with code for further training a video generating model on captured crop rows, so you can produce new synthetic scenes.
QUICK INFO
SOLUTION TYPE
Models, data & tools
TECHNOLOGY CATEGORY
Simulation, digital twins & world models
AREA OF WORK
Farm operations & automation
ACCESS
Free
DEPLOYMENT
Self-hosted / on-premise · Downloadable
SOURCE REVIEWED
August 23, 2026
What it helps with
- Make synthetic video of crop rows
- More training footage for field robots
- Train a video model on cotton rows
- Generate unusual field scenes for robot vision
- Recipe for further training a video model
- Tomato and cotton row footage for training
Scope & audience
WHO IT'S FOR
Developer / AI builder · Ag-tech product team
AREA OF WORK
Farm operations & automation
COMMODITY / SCOPE
Row crops · Cotton · Tomato
Access & technical details
ACCESS
Anyone can get it
INTERFACE
Code repository
DEPLOYMENT
You run it yourself
LANGUAGES
Not stated by provider
Featured media
Source information
This listing was compiled from publicly available source material. Verify current details with the provider.
PROVIDER DOCUMENTATION
Reviewed August 2026
Explore more like this
APSIM Next Generation
APSIM Initiative
Models, data & tools
You want to see how a crop might respond to different soils, weather or management before committing a field. This is a modular farming systems simulator you can set up for those scenarios and run yourself.
Paid or approval
Cloud-hosted
Simulation, digital twins & world models
APSIM Next Generation
plan-manage-production
agronomy-soil-and-crop-management
buildable-or-downloadable-component
simulation-digital-twins-and-world-models
Paid or approval
Cloud-hosted
August 17, 2026
- How will a crop respond to different weather
- Try a management change before committing a field
- Simulate crop growth on my soil
- Model farming system scenarios
- Test a rotation on paper
SynthDet
Unity Technologies
Models, data & tools
Hand labelling thousands of photos so a model can spot objects is dull, slow work. This is an end to end pipeline built on a game engine that renders images and their labels together, ready for training.
Free · Open source
Self-hosted / on-premise · Downloadable
Simulation, digital twins & world models
SynthDet
build-integrate-ai
agrifood-data-and-ai-infrastructure
buildable-or-downloadable-component
simulation-digital-twins-and-world-models
Free · Open source
Self-hosted / on-premise · Downloadable
August 17, 2026
- Make labelled images for object detection
- How do i avoid paying photo annotators
- Render training pictures with labels included
- Generate practice images in a game engine
- Train an object spotter without hand labelling
- End to end synthetic image pipeline
Is this your Resource?
Contact us to claim this listing and add details, screenshots or a demo.
