List view
- Look into the steps necessary to create a CI/CD pipeline - Learn the ins and outs of Docker containers and their usability - Test and verify the pipeline in different environments
Overdue by 4 year(s)•Due by June 19, 2021•2/2 issues closedBump up inferential performance of DeepGCN by: - looking into integrable modules (CRFs, interpolators, etc.) to achieve better fine-grained segmentation - Work with augmentation techniques - Look into optimizer choices + other hyperparameters
Overdue by 4 year(s)•Due by July 11, 2021•5/5 issues closedLook into fixes/tweaks which specifically deal with handling heavily class-imbalanced datasets. Work with loss functions, class-weighting, etc. Verify with "trial" dataset splits and visualize results with TB logs. If none work --> constrain target vector to more feasible and easily identifiable classes (Vegetation, Terrain, Road, Car, etc.)
Overdue by 4 year(s)•Due by May 21, 2021•2/2 issues closed