Preparing cleaned image counts.
Analysis Lab
Extensive EDA for the deployed research engine
This page summarizes the curated dataset, split protocol, optimization trace, leakage-control safeguards, and held-out evaluation context behind the deployed model.
Live analytics on this page are loaded from Loading....
Dataset Profile
Class balance and curation
Partition Protocol
Train, validation, and test composition
Loading split details...
Optimization Trace
Training dynamics across epochs
Training
Validation
Held-Out Evaluation
Current internal performance
Accuracy
Loading...
Loading evaluation scope...
Leakage Control
Similarity audit and duplicate handling
Loading audit signals...
Interpretation Notes
How to read this evidence responsibly
- Loading interpretation notes...
Distance Profile
Nearest-train similarity by evaluation split
Loading distance profile...
Artifact Registry
Saved audit files and data provenance
Loading artifact registry...