My code for classifying images from brodazt database with different transformations\distortions
All work organized by folders and subfolders, separated in different levels. Depth of folder is level. Folders which not related to next level has '_' prefix.
- Contains common data, imgs;
- Types of experiments (one against all, multi-classification, ...);
- Types of preprocessings (raw, standardize, tiling)
- Type of employed model, here we can see history of development of model, current results for tuning on train | valid.
111 classes of different textures. Each class represented with 5 grayscale images with size 200x200.
- comp1_test - shift??
- comp2_test - scale changed > 1
- comp3_test - scale changed < 1
- comp4_test - rotate