Sameera Ramasinghe
YOU?
Author Swipe
View article: DARB-Splatting: Generalizing Splatting with Decaying Anisotropic Radial Basis Functions
DARB-Splatting: Generalizing Splatting with Decaying Anisotropic Radial Basis Functions Open
Splatting-based 3D reconstruction methods have gained popularity with the advent of 3D Gaussian Splatting, efficiently synthesizing high-quality novel views. These methods commonly resort to using exponential family functions, such as the …
View article: Test-Time Adaptation of 3D Point Clouds via Denoising Diffusion Models
Test-Time Adaptation of 3D Point Clouds via Denoising Diffusion Models Open
Test-time adaptation (TTA) of 3D point clouds is crucial for mitigating discrepancies between training and testing samples in real-world scenarios, particularly when handling corrupted point clouds. LiDAR data, for instance, can be affecte…
View article: From Activation to Initialization: Scaling Insights for Optimizing Neural Fields
From Activation to Initialization: Scaling Insights for Optimizing Neural Fields Open
In the realm of computer vision, Neural Fields have gained prominence as a contemporary tool harnessing neural networks for signal representation. Despite the remarkable progress in adapting these networks to solve a variety of problems, t…
View article: Backpropagation-free Network for 3D Test-time Adaptation
Backpropagation-free Network for 3D Test-time Adaptation Open
Real-world systems often encounter new data over time, which leads to experiencing target domain shifts. Existing Test-Time Adaptation (TTA) methods tend to apply computationally heavy and memory-intensive backpropagation-based approaches …
View article: Robust Point Cloud Processing through Positional Embedding
Robust Point Cloud Processing through Positional Embedding Open
End-to-end trained per-point embeddings are an essential ingredient of any state-of-the-art 3D point cloud processing such as detection or alignment. Methods like PointNet, or the more recent point cloud transformer -- and its variants -- …
View article: A Learnable Radial Basis Positional Embedding for Coordinate-MLPs
A Learnable Radial Basis Positional Embedding for Coordinate-MLPs Open
We propose a novel method to enhance the performance of coordinate-MLPs (also referred to as neural fields) by learning instance-specific positional embeddings. End-to-end optimization of positional embedding parameters along with network …
View article: On the effectiveness of neural priors in modeling dynamical systems
On the effectiveness of neural priors in modeling dynamical systems Open
Modelling dynamical systems is an integral component for understanding the natural world. To this end, neural networks are becoming an increasingly popular candidate owing to their ability to learn complex functions from large amounts of d…
View article: Few-shot Class-incremental Learning for 3D Point Cloud Objects
Few-shot Class-incremental Learning for 3D Point Cloud Objects Open
Few-shot class-incremental learning (FSCIL) aims to incrementally fine-tune a model (trained on base classes) for a novel set of classes using a few examples without forgetting the previous training. Recent efforts address this problem pri…
View article: Trading Positional Complexity vs. Deepness in Coordinate Networks
Trading Positional Complexity vs. Deepness in Coordinate Networks Open
It is well noted that coordinate-based MLPs benefit -- in terms of preserving high-frequency information -- through the encoding of coordinate positions as an array of Fourier features. Hitherto, the rationale for the effectiveness of thes…
View article: Learning Positional Embeddings for Coordinate-MLPs
Learning Positional Embeddings for Coordinate-MLPs Open
We propose a novel method to enhance the performance of coordinate-MLPs by learning instance-specific positional embeddings. End-to-end optimization of positional embedding parameters along with network weights leads to poor generalization…
View article: Beyond Periodicity: Towards a Unifying Framework for Activations in Coordinate-MLPs
Beyond Periodicity: Towards a Unifying Framework for Activations in Coordinate-MLPs Open
Coordinate-MLPs are emerging as an effective tool for modeling multidimensional continuous signals, overcoming many drawbacks associated with discrete grid-based approximations. However, coordinate-MLPs with ReLU activations, in their rudi…
View article: Enabling equivariance for arbitrary Lie groups
Enabling equivariance for arbitrary Lie groups Open
Although provably robust to translational perturbations, convolutional neural networks (CNNs) are known to suffer from extreme performance degradation when presented at test time with more general geometric transformations of inputs. Recen…
View article: Rethinking Positional Encoding
Rethinking Positional Encoding Open
It is well noted that coordinate based MLPs benefit -- in terms of preserving high-frequency information -- through the encoding of coordinate positions as an array of Fourier features. Hitherto, the rationale for the effectiveness of thes…
View article: Robust normalizing flows using Bernstein-type polynomials
Robust normalizing flows using Bernstein-type polynomials Open
Modeling real-world distributions can often be challenging due to sample data that are subjected to perturbations, e.g., instrumentation errors, or added random noise. Since flow models are typically nonlinear algorithms, they amplify thes…
View article: Rethinking conditional GAN training: An approach using geometrically\n structured latent manifolds
Rethinking conditional GAN training: An approach using geometrically\n structured latent manifolds Open
Conditional GANs (cGAN), in their rudimentary form, suffer from critical\ndrawbacks such as the lack of diversity in generated outputs and distortion\nbetween the latent and output manifolds. Although efforts have been made to\nimprove res…
View article: Blended Convolution and Synthesis for Efficient Discrimination of 3D Shapes
Blended Convolution and Synthesis for Efficient Discrimination of 3D Shapes Open
Existing networks directly learn feature representations on 3D point clouds for shape analysis. We argue that 3D point clouds are highly redundant and hold irregular (permutation-invariant) structure, which makes it difficult to achieve in…
View article: Learned and Hand-crafted Feature Fusion in Unit Ball for 3D Object Classification
Learned and Hand-crafted Feature Fusion in Unit Ball for 3D Object Classification Open
Convolution is an effective technique that can be used to obtain abstract feature representations using hierarchical
\nlayers in deep networks. However, performing convolution in non-Euclidean topological spaces such as the unit
\nball (B
…
View article: Spectral-GANs for High-Resolution 3D Point-cloud Generation
Spectral-GANs for High-Resolution 3D Point-cloud Generation Open
Point-clouds are a popular choice for vision and graphics tasks due to their accurate shape description and direct acquisition from range-scanners. This demands the ability to synthesize and reconstruct high-quality point-clouds. Current d…
View article: Volumetric Convolution: Automatic Representation Learning in Unit Ball
Volumetric Convolution: Automatic Representation Learning in Unit Ball Open
Convolution is an efficient technique to obtain abstract feature representations using hierarchical layers in deep networks. Although performing convolution in Euclidean geometries is fairly straightforward, its extension to other topologi…
View article: A Context-aware Capsule Network for Multi-label Classification
A Context-aware Capsule Network for Multi-label Classification Open
Recently proposed Capsule Network is a brain inspired architecture that brings a new paradigm to deep learning by modelling input domain variations through vector based representations. Despite being a seminal contribution, CapsNet does no…