Neural Field Arena
收藏arXiv2023-12-17 更新2024-08-06 收录
下载链接：
http://arxiv.org/abs/2312.10531v1
下载链接
链接失效反馈官方服务：
资源简介：
Neural Field Arena是由阿姆斯特丹大学的研究团队开发的一个综合性基准，旨在评估和比较不同神经场（NeF）架构在多种视觉数据集上的表现。该基准包括了如MNIST、CIFAR、ImageNet变体及ShapeNetv2等数据集的神经场版本，用于标准化测试和推动神经场研究的进一步发展。通过这个基准，研究者可以系统地探索和优化神经场的超参数，如初始化、网络架构和优化策略，以提高其在下游任务中的性能。
Neural Field Arena is a comprehensive benchmark developed by a research team at the University of Amsterdam, aimed at evaluating and comparing the performance of different Neural Field (NeF) architectures across diverse visual datasets. This benchmark provides Neural Field-compatible versions of datasets including MNIST, CIFAR, ImageNet variants, and ShapeNetv2, which supports standardized testing and further advances research in Neural Fields. With this benchmark, researchers can systematically explore and optimize the hyperparameters of Neural Fields, such as initialization strategies, network architectures, and optimization methods, to improve their performance on downstream tasks.
提供机构：
阿姆斯特丹大学
创建时间：
2023-12-17
