Masters-object-detection

Masters 3D object detection implementation

LungAnnFix

• folder with code to convert segmentation annotation to bounding box (considering the bigger body as the desired nodule)

LungCoordinates

• folder with LUNA16 spacing and origin validation
• folder with scripts to convert voxel coordinates to world coordinates

LungCoordinatesValidation

• TODO: complete with info

LungDetectionTraining

• code to perform object 3D detection training

LungDetectionValidation

• code to perform object 3D detection validation

LungSegmentation

• code to perform lung segmentations (monai example, and the one that generated best results, with minimal time to process the data), with their visualization saving and mask saving. TODO: implement improvements to this algorithm!!
• contains reference code used by Lucas (in reference folder)

LungSegmentationValidation

• code used to verify if all the bounding box annotations fit the lung segmentation (TODO: need to finish implementation)

LungVisualizations

• TODO: move all visualization code to this folder!

POCS

reference

• contains KNN reference code

torch-utils

• contains example code, of how to run an object detection model, into a real image (2D)
• contains code to verify GPU usage

statistics

• contains code to generate histograms of the exams dimmensions

tutorial-to-object-detection

• contains tutorial to object detection (explanation of how it works) TODO: move to one note!

preprocessing -tutorial

• contains code to perform first version (vanila) of lung segmentation - execute fast, but return more errors
• contains code to perform watershed algorithm for lung segmentation (takes too long to generate files)
• contains a helper folder with examples to perform clahe, 3d-plot and matplitlib plot conversion to png

pedestrian-detection-mask-r-cnn

• mask-r-cnn example implementation, with pedestrian dataset (example from pytorch documentation, to perform object detection)