Classification Analysis Results

Results file: val-results.json

Ground truth file: val-gt.json

Detection threshold: 0.2

Classification confidence threshold: 0.5

Analysis level: sequence

Total sequences: 2487

Ground truth collapsed: single label per sequence

Predictions collapsed: single label per sequence

Accuracy	0.9260
Macro F1	0.8690
Micro F1	0.9260
Micro Precision	0.9260
Micro Recall	0.9260

Predicted as this category: images where this category was among the predictions, and some other category was among the true labels. This category may also be a correct true label for the image; these entries do not necessarily represent errors. For example, if the ground truth for an image is “human,cattle” and the prediction is “human” or “human,cattle”, this image would be included in the "human" link in this column in the “cattle” row. This column is generally useful for understanding the distribution of multi-label or multi-prediction images, but it is not very useful for understanding model errors.

Mispredicted as this category: images where this category was among the predictions, and it was wrong (not present in the ground truth). This column is useful for understanding model errors, specifically false positives for this category.

This was predicted as: images with this true label that also received a prediction of some other category. Other predictions (including the correct one) may also be present. For example, if the ground truth for an image is “cattle”, and the prediction is “human,cattle”, this image would be included in the "human" link in this column in the “cattle” row. This column is generally useful for understanding the distribution of multi-label or multi-prediction images, but it is not very useful for understanding model errors.

Mispredicted as: images with this true label whose predictions did not include this label. This column is useful for understanding model errors, specifically false negatives for this category.