Skip to main content

Description of compute-msi scripts

  • fastai_multi_label_v3_generic.py: Script to train a one-stage model or the second stage of a two-stage model.
  • fastai_multi_label_predict.py: Script to use the one-stage model to get predictions on test data (which is queried from the frames_test table).
  • nd_vs_defect_v1_generic.py: Script to train the first stage of a two-stage model.
  • cctv_predictions_binary_model.py: Script to use the first stage of the two-stage model to get predictions on test data.
  • cctv_predictions_filter_predicted_data.py: Don't know what this does.
  • cctv_multi_label_two_stage_approach_GQC_metric.py: Script to calculate the performance metrics for the two stage model.

GPU Usage and batch size correspondence of some compute_msi scripts:​

  • fastai_multi_label_v3_generic.py: A batch size of 12 corresponds to 8GB of GPU memory being allocated.
  • fastai_multi_label_predict.py: A batch size of 64 corresponds to 8GB of GPU memory being allocated.