Neural Field Classifiers via Target Encoding and Classification Loss

Authors: Xindi Yang, Zeke Xie, Xiong Zhou, Boyu Liu, Buhua Liu, Yi Liu, Haoran Wang, YUNFENG CAI, Mingming Sun

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments demonstrate the impressive effectiveness of NFC at the nearly free extra computational costs. Moreover, NFC also shows robustness to sparse inputs, corrupted images, and dynamic scenes.
Researcher Affiliation Collaboration µBeijing Key Lab of Traffic Data Analysis and Mining, Beijing Jiaotong University γCognitive Computing Lab, Baidu Research
Pseudocode No The paper describes the methodology using text and diagrams (Figure 1 and 2), but it does not include any explicitly labeled pseudocode or algorithm blocks.
Open Source Code Yes Code: https://github.com/Madaoer/Neural-Field-Classifier.
Open Datasets Yes We choose the Replica Dataset and Tanks and Temples Advanced(T & T) as the benchmark datasets. The Replica Dataset encompasses eight complex static scenes... T & T Dataset is a popular 3D reconstruction dataset... We also use the classical Chair scene for evaluating NFC and NFR with Strivec (Gao et al., 2023).
Dataset Splits Yes In the experiments on Replica, if one image index is divisible by 10, we move the image to the test dataset; if not, we move the image to the training dataset. Thus, we have 90% images for training and 10% images for evaluation.
Hardware Specification Yes Table 7: Ablation study I. Model: DVGO. Dataset: Replica Scene 6. GPU: A100. Table 8: Ablation study II. Model: Neu S. Dataset: Replica Scene 7. GPU: A100.
Software Dependencies No The paper mentions software like 'Adam' (an optimizer), 'Nerf-pytorch' (implied from citation), and 'SDFStudio' but does not provide specific version numbers for these software components.
Experiment Setup Yes DVGO Setting We employ the sourcecode of DVGO (Version 2) in the original project (Sun et al., 2022) without modifying training hyperparameters. So we train DVGO via Adam (Kingma and Ba, 2015) with the batch size B = 4096. The learning rate of density voxel grids and color/feature voxel grids is 0.1, and the learning rate of the RGB net (MLP) is 0.001. The total number of iterations is 30000. We multiply the learning rate by 0.1 per 1000 iterations.