mlreco.post_processing.analysis.instance_clustering module

mlreco.post_processing.analysis.instance_clustering.instance_clustering(cfg, data_blob, res, logdir, iteration)[source]

Simple DBSCAN on uresnet clustering output for instance segmentation

Parameters

  • data_blob (dict) – Input dictionary returned by iotools

  • res (dict) – Results from the network, dictionary using analysis_keys

  • cfg (dict) – Configuration

  • idx (int) – Iteration number

  • Input –

  • ----- –

  • the following analysis keys (Requires) –

  • segmentation (-) –

  • clustering (-) –

  • the following data blob keys (Requires) –

  • input_data (-) –

  • segment_label UResNet 5 classes label (-) –

  • cluster_label (-) –

Output

  • Writes 2 CSV files

  • - `instance_clustering-*` with the clustering predictions (point type 0 =

  • event data, point type 1 = predictions, point type 2 = T-SNE visualizations)

  • - `instance_clustering_metrics-*` with some event-wise metrics such as AMI and ARI.

mlreco.post_processing.analysis package mlreco.post_processing.analysis.michel_reconstruction module