Learning to Decompose and Disentangle Representations for Video Prediction

Authors: Jun-Ting Hsieh, Bingbin Liu, De-An Huang, Li F. Fei-Fei, Juan Carlos Niebles

NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate DDPAE on two datasets: Moving MNIST [31] and Bouncing Balls [3].Our goal is to predict a sequence of future frames given a sequence of input frames.
Researcher Affiliation Academia Jun-Ting Hsieh Stanford University junting@stanford.eduBingbin Liu Stanford University bingbin@stanford.eduDe-An Huang Stanford University dahuang@cs.stanford.eduLi Fei-Fei Stanford University feifeili@cs.stanford.eduJuan Carlos Niebles Stanford University jniebles@cs.stanford.edu
Pseudocode No The paper describes the model implementation and learning process in prose and figures, but does not include any explicit pseudocode or algorithm blocks.
Open Source Code Yes Code for DDPAE and the experiments are available at https://github.com/jthsieh/DDPAE-video-prediction.
Open Datasets Yes Moving MNIST has been widely used for evaluating video prediction models [15, 31, 43].The test set is a fixed dataset downloaded from [31] consisting of 10,000 sequences of 20 frames, with 10 as input and 10 to predict.We simulate sequences of 4 balls bouncing in an image with the physics engine code used in [3].
Dataset Splits No The paper specifies training and test set sizes for Bouncing Balls ('We generated a fixed training set of 50,000 sequences and a test set of 2,000 sequences.') and describes the test set for Moving MNIST ('The test set is a fixed dataset downloaded from [31] consisting of 10,000 sequences of 20 frames, with 10 as input and 10 to predict.'), but it does not explicitly define a separate validation dataset split.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., GPU models, CPU types, memory) used for running the experiments.
Software Dependencies No The paper mentions using a 'physics engine code used in [3]' and 'seq2seq based model', but does not specify software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions, or specific library versions).
Experiment Setup No The paper describes some aspects of dataset generation (e.g., 'randomly sampled velocity and angle' for Moving MNIST, 'maximum velocity is 60 pixels/second' for Bouncing Balls), but it does not explicitly provide concrete hyperparameter values or detailed system-level training configurations for the model itself.