Back to search
Deep Learning CNN using FastAI for the Stanford MRNet Knee MRI diagnosis challenge
Stars
16
Forks
8
Watchers
16
Open Issues
0
Overall repository health assessment
No package.json found
This might not be a Node.js project
tweaking to run baseline model; first pass prototype (alexnet) on mid-slice-only data
8f9ffc6View on GitHubminor adjustments to work more directly with dir structure of downloaded MRNet data files.
95be3f8View on GitHubcode for running image-based classifiers like AlexNet (original paper) and ResNet on slices from the scans, collecting ave AUC and other metrics
faf7d72View on GitHubbatch preprocessing step to save one or three slices from each scan as an image file
4383816View on GitHubexample notebook for using MRNetCaseList, a custom ItemList to take advantage of fastai DataBlock API
deda604View on GitHubinterpolate input data, which have varying numbers of slices, so that all scans have same numbers of slices
11dae8dView on GitHubadded custom show_xys for custom show_batch functionality
077c2a5View on GitHubchange MRNetCase .data attribute to numpy array of scans from all three planes; make reconstruct convert the corresponding tensor back into tuple of arrays, one for each plane
d2a70a5View on GitHubdefine a method for associating a df with labels associated with each Case in the CaseList; assign value to CaseList.inner_df
ef7fd9aView on GitHub