WateRF: Robust Watermarks in Radiance Fields for Protection of Copyrights
Authors:
Youngdong Jang,
Dong In Lee,
MinHyuk Jang,
Jong Wook Kim,
Feng Yang,
Sangpil Kim
Abstract:
The advances in the Neural Radiance Fields (NeRF) research offer extensive applications in diverse domains, but protecting their copyrights has not yet been researched in depth. Recently, NeRF watermarking has been considered one of the pivotal solutions for safely deploying NeRF-based 3D representations. However, existing methods are designed to apply only to implicit or explicit NeRF representat…
▽ More
The advances in the Neural Radiance Fields (NeRF) research offer extensive applications in diverse domains, but protecting their copyrights has not yet been researched in depth. Recently, NeRF watermarking has been considered one of the pivotal solutions for safely deploying NeRF-based 3D representations. However, existing methods are designed to apply only to implicit or explicit NeRF representations. In this work, we introduce an innovative watermarking method that can be employed in both representations of NeRF. This is achieved by fine-tuning NeRF to embed binary messages in the rendering process. In detail, we propose utilizing the discrete wavelet transform in the NeRF space for watermarking. Furthermore, we adopt a deferred back-propagation technique and introduce a combination with the patch-wise loss to improve rendering quality and bit accuracy with minimum trade-offs. We evaluate our method in three different aspects: capacity, invisibility, and robustness of the embedded watermarks in the 2D-rendered images. Our method achieves state-of-the-art performance with faster training speed over the compared state-of-the-art methods.
△ Less
Submitted 27 May, 2024; v1 submitted 3 May, 2024;
originally announced May 2024.
Dynamical prediction of two meteorological factors using the deep neural network and the long short term memory $(1)$
Authors:
Ki Hong Shin,
Jae Won Jung,
Sung Kyu Seo,
Cheol Hwan You,
Dong In Lee,
Jisun Lee,
Ki Ho Chang,
Woon Seon Jung,
Kyungsik Kim
Abstract:
It is important to calculate and analyze temperature and humidity prediction accuracies among quantitative meteorological forecasting. This study manipulates the extant neural network methods to foster the predictive accuracy. To achieve such tasks, we analyze and explore the predictive accuracy and performance in the neural networks using two combined meteorological factors (temperature and humid…
▽ More
It is important to calculate and analyze temperature and humidity prediction accuracies among quantitative meteorological forecasting. This study manipulates the extant neural network methods to foster the predictive accuracy. To achieve such tasks, we analyze and explore the predictive accuracy and performance in the neural networks using two combined meteorological factors (temperature and humidity). Simulated studies are performed by applying the artificial neural network (ANN), deep neural network (DNN), extreme learning machine (ELM), long short-term memory (LSTM), and long short-term memory with peephole connections (LSTM-PC) machine learning methods, and the accurate prediction value are compared to that obtained from each other methods. Data are extracted from low frequency time-series of ten metropolitan cities of South Korea from March 2014 to February 2020 to validate our observations. To test the robustness of methods, the error of LSTM is found to outperform that of the other four methods in predictive accuracy. Particularly, as testing results, the temperature prediction of LSTM in summer in Tongyeong has a root mean squared error (RMSE) value of 0.866 lower than that of other neural network methods, while the mean absolute percentage error (MAPE) value of LSTM for humidity prediction is 5.525 in summer in Mokpo, significantly better than other metropolitan cities.
△ Less
Submitted 16 January, 2021;
originally announced January 2021.
