Table 3 Summary of training and validation of the DeepLabCut models.

From: Multimodal dataset linking wide‐field calcium imaging to behavior changes in operant lever‐pull task in mice

 

Training and testing

Validation

Extracted

Annotated

Annotated

Animals

Sessions

Images

Animals

Sessions

Images

Upper body

19 + 4

35 + 8

1360 + 40

1399

25

96

192

Face

6 + 0

12 + 0

240 + 0

240

25

96

192

Eye

23 + 4

39 + 8

720 + 40

676

25

91

180

  1. Training of the models was performed incrementally, through checking video files one-by-one. The numbers of animals, sessions, and images for training and testing (left) corresponds to the numbers from the videos of the current dataset, whereas those for validation (right) indicate the numbers from different experiments in the same behavioral rig, with the same video-camera configurations. For more details of our testing and validation strategy, see Fig. S3. For the eye model, in particular, discrepancy between the numbers of extracted and annotated frames occurs due to extraction of images when the animals were blinking.
Back to article page