CudaVision

Learning Vision systems on Graphics cards.

Assignment 1 : Logistic Regression on CIFAR-10 dataset using pure numpy.

MNIST with MLP : Multilayer Perceptron performed on MNIST dataset.

Assignment : CIFAR10 - Using Logistic Regression. (Assignment 4)
(I) Add regularization to your CFAR-10 model (Dropout, L1, and L2).
(II) Find how deep the model can be and still improving.
(III)Draw confusion matrix.
(IV) Use five different optimizers and compare their results and convergence time (SGD, Adam, Adagrad, Adadelta, RMSprop).
(V) Use three different activation functions (ReLu, Tanh, Sigmoid)

Assignment 2 : Convolutional Neural Network (CNN).

Assignment 3 : Convolutional Autoencoder on CIFAR10.

Assignment 5 : To fine tune a pretrained network.
Dataset: https://drive.google.com/file/d/1I4928zz9hu02TNPQ2fUN-gJN-vJsxH2K/view

Assignment 7 : LSTM and a GRU model for MNIST dataset

Assignment 8 : Generator - Descriminator model for the soccerball dataset.

Project : Detection and Localization of a Soccer ball from an input image with feed-forward Fully Convolutional Neural Networks (FCNN).
(I) To implement SweatyNet1 model from Homonoids Robocup workshop 2017 for detection and localization of Soccer ball.
(II) Train a convLSTM model on top of it with continuous sequence of frames.
(III) Calculate recall and false detection rate to check accuracy of model.
Dataset: https://drive.google.com/open?id=1Kua5N1xYUMfGIr_mClAejdW2BzvTrDUR