Radio source-component association for the LOFAR Two-metre Sky Survey with region-based convolutional neural networks

Open Access

Table 1

Maximum catalogue accuracy attained by different CNN backbones on the large and bright source components.

CNN backbone	Train accuracy	Val. accuracy
ResNet 50 layers	89.0% ± 0.3	84.3% ± 0.4
ResNeXt 50 layers	87.7% ± 0.3	82.8% ± 0.2
ResNet 101 layers	88.5% ± 0.2	83.1% ± 0.9
ResNeXt 101 layers	87.7% ± 0.5	82.8% ± 0.5

Notes. The reported numbers in this and further tables are the mean and standard deviations of three training runs with different random seeds and otherwise equal set-ups.

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.