Batch reports from specimen images.
Saved benchmark results. Browse completed predictions; opening this page does not run inference.
Model and validation context
Model and report source
Loading saved predictions and batch reports...
Execution and evaluation scope
Predictions were computed before this page was opened.
Estimated label-group composition
Complete JSON ↓Top-label counts assign each specimen to one group. Probability-weighted counts spread its contribution across groups.
BATCH CONTEXT
—
Probability-based—
Low-confidence predictions
How to interpret this report
Low-confidence predictions are flagged automatically. Small probabilities can accumulate and inflate estimated presence. These are label-group estimates, not confirmed species inventories.
Reports cover imaged specimens in the official test split, not an entire waterbody. They do not assess drinking-water safety, ecological status, or human health.
Imaged specimens only · Label-group estimates
Browse a sample of specimens with anonymous IDs. Changing views does not run inference or alter saved predictions.
Image index
Model predictions
Saved specimen-level probabilities