GRAF: Generative Radiance Fields for 3D-Aware Image Synthesis
Authors: Katja Schwarz, Yiyi Liao, Michael Niemeyer, Andreas Geiger
NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We systematically analyze our approach on several challenging synthetic and real-world datasets. Our experiments reveal that radiance fields are a powerful representation for generative image synthesis, leading to 3D consistent models that render with high fidelity. |
| Researcher Affiliation | Academia | Katja Schwarz Yiyi Liao Michael Niemeyer Andreas Geiger Autonomous Vision Group MPI for Intelligent Systems and University of Tübingen {firstname.lastname}@tue.mpg.de |
| Pseudocode | No | The paper describes the model architecture and training procedures in text and with diagrams (e.g., Figure 2), but it does not include any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | Yes | We release our code and datasets at https://github.com/autonomousvision/graf. |
| Open Datasets | Yes | We consider two synthetic and three real-world datasets in our experiments. ... We render 150k Chairs from Photoshapes [49]... We further use the Carla Driving simulator [12] to create 10k images... We use the Faces dataset which comprises celeb A [31] and celeb A-HQ [24]... In addition, we consider the Cats dataset [73] and the Caltech-UCSD Birds-200-2011 [66] dataset. |
| Dataset Splits | No | The paper mentions using datasets for training and evaluation but does not specify the exact percentages or methods for creating training, validation, and test splits (e.g., 80/10/10 split or specific sample counts for each split). |
| Hardware Specification | No | The paper does not provide specific details about the hardware used for running the experiments, such as GPU models, CPU specifications, or memory. |
| Software Dependencies | No | The paper mentions algorithms and techniques like "RMSprop [27]", "spectral normalization [37]", and "instance normalization [65]", but it does not specify any software libraries, frameworks, or their version numbers (e.g., Python, PyTorch, TensorFlow, CUDA versions). |
| Experiment Setup | Yes | We use spectral normalization [37] and instance normalization [65] in our discriminator and train our approach using RMSprop [27] with a batch size of 8 and a learning rate of 0.0005 and 0.0001 for generator and discriminator, respectively. At inference, we randomly sample zs, za and ξ, and predict a color value for all pixels in the image. |