Self-supervised Video Prediction Project

We use the DSSIM and L2 loss for training. We use three seed frames and predict the next three frames in an autoregressive manner:

The results of the model on the Moving MNIST testing dataset (http://www.cs.toronto.edu/~nitish/unsupervised_video/) can be seen below :

Results on a customized Robot dataset (created from scratch from youtube videos) are visible below:

Other Deep Learning models

DL algorithms implemented using Pytorch (on CUDA GPUs):

List of algorithms:

Name		Name	Last commit message	Last commit date
Latest commit History 7 Commits
Images		Images
PyTorch_CUDA_CNN_CIFAR10.ipynb		PyTorch_CUDA_CNN_CIFAR10.ipynb
PyTorch_CUDA_DCGAN.ipynb		PyTorch_CUDA_DCGAN.ipynb
PyTorch_CUDA_HyperparameterSweeps_WandB.ipynb		PyTorch_CUDA_HyperparameterSweeps_WandB.ipynb
PyTorch_CUDA_LSTM_GRU.ipynb		PyTorch_CUDA_LSTM_GRU.ipynb
PyTorch_CUDA_RobotDataset.ipynb		PyTorch_CUDA_RobotDataset.ipynb
PyTorch_CUDA_VariationalConvolutionalAutoencoder.ipynb		PyTorch_CUDA_VariationalConvolutionalAutoencoder.ipynb
PyTorch_CUDA_VisualizingActivations.ipynb		PyTorch_CUDA_VisualizingActivations.ipynb
PyTorch_CUDA_WGAN.ipynb		PyTorch_CUDA_WGAN.ipynb
README.md		README.md