Tensorimage

New features:
- Live classification: classify images in real time without storing predictions on disk
- Live training: update a model on real time with new labeled data being constantly fed to a trainer
- Data augmentation operations:
- Affine transformations
- Perspective transformations
- Piecewise affine transformations
- Image translations
- Median blurring
- Bilateral blurring
- Motion blurring
- Grayscale filtering
- Element wise subtraction, division, addition and multiplication
- Image gamma adjusting

New features:
* Cluster training:
* Useful for feature engineering and comparing performance of different trainers' with different hyperparameters
* Compare an unlimited amount of trainers' performance based on testing accuracy and run trainers in parallel in CPUs (limited to the number of CPUs available)

TensorImage is now a library, meaning that it can now be used for a wider range of purposes than in previous versions, where it could only be used as a CLI tool. New features:
* Data augmentation:
* Image flipping
* Salt pepper noise
* Lighting modification
* Image rotation
* CNN architectures:
* RosNet
* AlexNet