Welcome to the DeepLabCut Benchmark!#

This page hosts the official DeepLabCut Benchmark.

Leaderboard#

Click on columns to sort in ascending or descending order. Press Ctrl while selecting columns to sort by multiple columns. Hover over a column name to get additional explanations for the metric.

TRI-MOUSE#

Model Dataset RMSE (assembly) RMSE (detections) mAP
- DLCRNet_ms4 (30k) TRI-MOUSE 3.55 3.567 0.947
- EfficientNet B7_s4 (30k) TRI-MOUSE 4.93 4.939 0.959
- ResNet50 (30k) TRI-MOUSE 3.97 4.216 0.943

FISH#

Model Dataset RMSE (assembly) RMSE (detections) mAP
- DLCRNet_ms4 (30k) FISH 7.15 4.918 0.75
- EfficientNet B7_s4 (30k) FISH 6.85 4.463 0.75
- ResNet50_s4 (30k) FISH 7.03 4.36 0.736

MARMOSETS#

Model Dataset RMSE (assembly) RMSE (detections) mAP
- DLCRNet (200k) MARMOSETS 8.1 8.815 0.791
- DLCRNet_ms4 (200k) MARMOSETS 7.56 8.504 0.808
- EfficientNet B7 (200k) MARMOSETS 7.27 7.453 0.818
- EfficientNet B7_s4 (200k) MARMOSETS 6.83 7.042 0.817

PARENTING-MOUSE#

Model Dataset RMSE (assembly) RMSE (detections) mAP
- DLCRNet_ms4 (30k) PARENTING 9.85 10.026 0.629
- EfficientNet B7 (30k) PARENTING 10.24 8.611 0.627
- EfficientNet B7_s4 (30k) PARENTING 8.9 9.355 0.639