Table 4
Results on the large and bright source component test dataset using our trained model.
| Total | All components 1121 (100.0%) | Single component source 701 (62.5%) | Part of a multi-component source 420 (37.5%) |
|---|---|---|---|
| Prediction is correct (TP) | 962 (85.8%) | 650 (58.0%) | 312 (27.8%) |
| Prediction includes too few components (FP+FN) | 63 (5.6%) | 1 (0.1%) | 62 (5.5%) |
| Prediction includes too many components (FP) | 96 (8.6%) | 54 (4.8%) | 42 (3.7%) |
Notes. The number of components in each category are also given as a percentage of the total number of components in the test dataset. The false positives (FPs) are broken down further into two categories. If the prediction for a component simultaneously includes both too many and too few components, it is grouped in the ‘too many’ category.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.