Learning Spatially Collaged Fourier Bases for Implicit Neural Representation

Authors: Jason Chun Lok Li, Chang Liu, Binxiao Huang, Ngai Wong

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Comprehensive experiments demonstrate the superior reconstruction quality of the proposed approach over existing baselines across various INR tasks, including image fitting, video representation, and 3D shape representation.
Researcher Affiliation Academia Department of Electrical and Electronic Engineering, The University of Hong Kong {jasonlcl, lcon7, huangbx7}@connect.hku.hk, nwong@eee.hku.hk
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code No The paper mentions that baselines are 'based on official codes released by authors of the respective models' but does not provide a statement or link for the open-sourcing of SCONE's code.
Open Datasets Yes The image representation task is performed on selected images from the Kodak dataset (Eastman Kodak Company 1999)... For the 3D shape representation task, we use the Stanford 3D scan dataset.
Dataset Splits No The paper describes sampling data for training but does not provide specific details on train/validation/test dataset splits, percentages, or explicit counts needed for reproduction.
Hardware Specification Yes All models are trained for 10k iterations on Nvidia RTX 3090 GPUs, each equipped with a 24GB memory buffer.
Software Dependencies No The implementation of all codes is carried out using the Py Torch (Paszke et al. 2019) framework, but no specific version number for PyTorch or other software dependencies is provided.
Experiment Setup Yes All models are trained for 10k iterations on Nvidia RTX 3090 GPUs, each equipped with a 24GB memory buffer. The training process utilizes the Adam optimizer (Kingma and Ba 2014), with the parameters β1 = 0.9 and β2 = 0.999, and employs the Mean Squared Error (MSE) loss without weight decay. Additionally, a cosine learning rate scheduler is applied, with a minimum learning rate of 1e 6.