3D-Aware Scene Manipulation via Inverse Graphics

Authors: Shunyu Yao, Tzu Ming Hsu, Jun-Yan Zhu, Jiajun Wu, Antonio Torralba, Bill Freeman, Josh Tenenbaum

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments demonstrate that our editing scheme based on 3D-SDN is superior to its 2D counterpart.
Researcher Affiliation Collaboration Shunyu Yao IIIS, Tsinghua University Tzu-Ming Harry Hsu MIT CSAIL Jun-Yan Zhu MIT CSAIL Jiajun Wu MIT CSAIL Antonio Torralba MIT CSAIL William T. Freeman MIT CSAIL, Google Research Joshua B. Tenenbaum MIT CSAIL
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes Please check out our code and website for more details. ... The code and full results can be found at our website.
Open Datasets Yes We conduct experiments on two street scene datasets: Virtual KITTI [Gaidon et al., 2016] and Cityscapes [Cordts et al., 2016].
Dataset Splits No The paper specifies training and testing splits but does not explicitly mention a separate validation split for reproducing experiments. It states: 'For each world, we use either the first or the last 80% consecutive frames for training and the rest for testing.'
Hardware Specification No The paper does not provide specific details about the hardware used for experiments.
Software Dependencies No The paper mentions software tools like Mask-RCNN and VGG network but does not provide specific version numbers for software dependencies.
Experiment Setup Yes We empirically set λreproj = 0.1. We first train the network with Lpred using Adam [Kingma and Ba, 2015] with a learning rate of 10^-3 for 256 epochs and then fine-tune the model with Lpred + λreproj Lreproj and REINFORCE with a learning rate of 10^-4 for another 64 epochs. ... We set λFM = 5 and λRecon = 10, and train the textural branch for 60 epochs on Virtual KITTI and 100 epochs on Cityscapes.