![]() ![]() The model was trained on the KITTI dataset. Next frame predictions on the Caltech Pedestrian dataset are shown below. ![]() An animation of the flow of information in the network can be found here. ![]() The PredNet architecture is illustrated below. It is trained for next-frame video prediction with the belief that prediction is an effective objective for unsupervised (or "self-supervised") learning. The PredNet is a deep convolutional recurrent neural network inspired by the principles of predictive coding from the neuroscience literature. PredNet by coxlab PredNet Code and models accompanying "Deep Predictive Coding Networks for Video Prediction and Unsupervised Learning" Bill Lotter, Gabriel Kreiman, and David Cox (2016) View on arXiv View on GitHub Download. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |