Time-Agnostic Prediction: Predicting Predictable Video Frames

Authors: Dinesh Jayaraman, Frederik Ebert, Alexei Efros, Sergey Levine

ICLR 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate our approach for future and intermediate frame prediction across three robotic manipulation tasks.
Researcher Affiliation Academia Dinesh Jayaraman UC Berkeley Frederik Ebert UC Berkeley Alyosha Efros UC Berkeley Sergey Levine UC Berkeley
Pseudocode No The paper describes network architectures and loss functions in detail but does not include any pseudocode or algorithm blocks.
Open Source Code No Note that more supplementary material, such as video examples, is hosted at: https://sites.google.com/view/ ta-pred (The provided URL links to video examples and related work, but not the source code for this specific paper's methodology or implementation).
Open Datasets Yes Finally, we test on BAIR pushing (Ebert et al., 2017), a real-world dataset that is commonly used in visual prediction tasks.
Dataset Splits No 5% of the data is set aside for testing. (No explicit mention of a separate validation split or the percentage for training data).
Hardware Specification Yes The NVIDIA DGX-1 used for this research was donated by the NVIDIA Corporation.
Software Dependencies No The paper mentions software like MuJoCo for data generation and implicitly uses PyTorch (via references to DCGAN architecture in Appendix C) but does not provide specific version numbers for any software dependencies.
Experiment Setup Yes For this pretraining, we use learning rate 0.0001 for 10 epochs with batch size 64 and Adam optimizer. Thereafter, for training, we use learning rate 0.0001 for 200 epochs with batch size 64 and Adam optimizer.