Neural ordinary differential equations (NODE) have been proposed as a continuous depth generaliza... more Neural ordinary differential equations (NODE) have been proposed as a continuous depth generalization to popular deep learning models such as Residual networks (ResNets). They provide parameter efficiency and automate the model selection process in deep learning models to some extent. However, they lack the much-required uncertainty modelling and robustness capabilities which are crucial for their use in several real-world applications such as autonomous driving and healthcare. We propose a novel and unique approach to model uncertainty in NODE by considering a distribution over the end-time T of the ODE solver. The proposed approach, latent time NODE (LT-NODE), treats T as a latent variable and apply Bayesian learning to obtain a posterior distribution over T from the data. In particular, we use variational inference to learn an approximate posterior and the model parameters. Prediction is done by considering the NODE representations from different samples of the posterior and can ...
In this paper a new video-based interface for creating cutout-style animation using magnets is pr... more In this paper a new video-based interface for creating cutout-style animation using magnets is presented. This idea makes room for users of all skill levels to animate. We created an interface which is generally a closed box with a camera placed at the bottom. Roof of the box acts like a multi touch interface. A cast of physical characters are designed by the animator using paper, markers and scissor. If animator wants to animate using this physical characters, he pastes them to magnets (we call them as PMagnets) and place them under the roof (puppets facing the camera). These characters are controlled by other magnets (CMagnets) on the top of the roof. Here we require, a simple foreground extraction algorithm to extract the characters and to render them onto a new background. Our system “Cut-Out Animation using Magnet Motion ” runs in real time (i.e 30 Frames/Sec). Therefore animator and the audience can instantly see the animation.
Neural ordinary differential equations (NODE) have been proposed as a continuous depth generaliza... more Neural ordinary differential equations (NODE) have been proposed as a continuous depth generalization to popular deep learning models such as Residual networks (ResNets). They provide parameter efficiency and automate the model selection process in deep learning models to some extent. However, they lack the much-required uncertainty modelling and robustness capabilities which are crucial for their use in several real-world applications such as autonomous driving and healthcare. We propose a novel and unique approach to model uncertainty in NODE by considering a distribution over the end-time T of the ODE solver. The proposed approach, latent time NODE (LT-NODE), treats T as a latent variable and apply Bayesian learning to obtain a posterior distribution over T from the data. In particular, we use variational inference to learn an approximate posterior and the model parameters. Prediction is done by considering the NODE representations from different samples of the posterior and can ...
In this paper a new video-based interface for creating cutout-style animation using magnets is pr... more In this paper a new video-based interface for creating cutout-style animation using magnets is presented. This idea makes room for users of all skill levels to animate. We created an interface which is generally a closed box with a camera placed at the bottom. Roof of the box acts like a multi touch interface. A cast of physical characters are designed by the animator using paper, markers and scissor. If animator wants to animate using this physical characters, he pastes them to magnets (we call them as PMagnets) and place them under the roof (puppets facing the camera). These characters are controlled by other magnets (CMagnets) on the top of the roof. Here we require, a simple foreground extraction algorithm to extract the characters and to render them onto a new background. Our system “Cut-Out Animation using Magnet Motion ” runs in real time (i.e 30 Frames/Sec). Therefore animator and the audience can instantly see the animation.
Uploads
Papers by Srinivas Anumasa