PyTorch Utility Package to setup training and testing pipeline for Computer Vision Tasks
Package has 5 sub-packages
Consists of Dataset, Dataloader functions and classes. Has a custom dataset class, along with transforms, gradcam visualization etc.
Includes two different network files, based on CIFAR-10 and MNIST.
Consists of Train and Testing part of NeuralNet. Mainly 3 functions, train, test and run. Requires Model and Modelconfig to be sent as input.
Mainly modelsummary with Receptive Field calculated layer-wise.
Consists of DataUtils and ModelUtils, which has helper functions mainly to plot and visualize data in former, & latter has model related functions.
- Normal 2d Convolutions
- Depthwise
- Dilated
- BatchNorm
- GroupNorm
- LayerNorm
- with layer-wise Receptive Field
- Cross Entropy Loss
- NLLoss
* Accuracy
* Stochastic Gradient Descent
* Step LR
* Reduce LR on Plateau
* One Cycle Policy
- MNIST
- CIFAR10