top of page

- 
Plot Feature For Multi Model Training
​
- 
Support for Imbalanced Datasets (simple and fast handling for unbalanced datasets)
​
- 
Model Compression
​
- 
Custom TFRecords Formats
​
- 
Custom Model Creation
​
- 
Improvisation in training loop​​
​
- 
Test Time Augmentations
​
- 
SOTA Augmentations
​​
- 
Label Smoothing
​
- 
Smoother training with option of batch wise or epoch wise LR update.
​
- 
Addition of Logging facilities
​
- 
GRADCAM for model output visualization & debugging
​
- 
Support for ResNext, SeResNet, SeResNext, NasNet (all variants).
​
- 
Wide range of evaluation metrics.
​
​
bottom of page
