ContainersNVIDIADeepSeek-R1NVIDIALlama-3.1-Nemotron-70B-InstructNVIDIAPyTorch
CollectionsNVIDIAOmniverse Kit (FB)NVIDIADeepStream SDKNVIDIAOmniverse Kit App Streaming
ModelsNVIDIAStyleGAN3 pretrained modelsNVIDIAPeopleNetNVIDIATrafficCamNet
ResourcesNVIDIARiva Skills Quick Start
Helm ChartsNVIDIAGPU OperatorNVIDIANVIDIA NIM Operator
NVIDIA
NVIDIA
CWIP (Contrastive World-Image Pre-training)
Model
NVIDIA
NVIDIA
CWIP (Contrastive World-Image Pre-training)

CWIP (Contrastive World-Image Pre-training) projects a camera frame and a world-model frame into a joint embedding space to score camera-to-world consistency and emit per-patch object-presence and object-type classifications.

This model is backed by NVIDIA's Plus Plus (++) Promise
to learn more about the quality of the datasets used to train this model.
FieldResponse
Intended Task/Domain:Synthetic Data Quality Evaluation
Model Type:Transformer
Intended Users:Developers and researchers building, evaluating, or auditing generative world foundation models for autonomous driving, including users of NVIDIA Cosmos and the open-source Cosmos Evaluator suite. Specific applications include per-frame embeddings, alignment scoring, per-patch defect detection, object-type classification, and automated quality scoring of driving-video generations.
Output:L2-normalized 1024-dimensional global embedding vectors for world-model frame and camera frame, and segmentation masks for per-patch defect classifications (MATCH, EXTRA, MISSING, WRONG, DONT_CARE) and object-type classifications over an 18-class taxonomy respectively.
Describe how the model works:CWIP projects a camera frame and a world-model frame into a joint embedding space to score camera-to-world consistency and emit per-patch object-presence and object-type classifications. It uses a transformer-based architecture to process visual inputs and generate embeddings and segmentation masks for quality scoring of driving-video generations.
Name the adversely impacted groups this has been tested to deliver comparable outcomes regardless of:Not Applicable
Technical Limitations and Mitigation:This model may not detect all inconsistencies between world-model control frame and camera frame. This model is trained on pairs of world-model control frame and camera frame. For other control signals, adaptation to the world-model control format or re-training with those specific control signals may be needed.
Verified to have met prescribed NVIDIA quality standards:Yes
Performance Metrics:Precision, Recall, F1, and IoU
Potential Known Risks:This model may output inaccurate semantic segmentation or suboptimal global embeddings that lead to incorrect synthetic image/video defect detection.
Terms of Use/Licensing:OpenMDW-1.1

NVIDIA uses cookies to improve your experience on our web site. We and our third-party partners also use cookies and other tools to collect and record information you provide as well as information about your interactions with our websites for performance improvement, analytics, and to assist in marketing efforts. By clicking "Accept All", you consent to our use of cookies and other tools as described in our Cookie Policy. You can manage your cookie settings by clicking on "Manage Settings." By continuing to use this site or by clicking one of the buttons below, you agree to our Terms of Service (which contains important waivers). Please see our Privacy Policy for more information on our privacy practices.