Paper-Conference

One limitation of existing Transformer-based models is that they cannot handle very long sequences as input since their self-attention …

Given a set of images from different views and their corresponding camera poses, PAPR learns a point-based surface representation of the scene and a rendering pipeline from scratch. Additionally, PAPR enables practical applications such as geometry editing, object manipulation, texture transfer, and exposure control.

We revisit the quadrature used to approximate volume rendering in NeRF and devise a different quadrature formula based on a different approximation to the opacity. We derive the rendering equation under piecewise linear opacity and show that it both resolves quadrature instability and has good numerical conditioning.

Our approach tackles the task of controllable part-based point cloud generation, where we are able to both generate novel shapes and perform edits on user-input shapes. An overview of our network is given at the bottom row. We factorize the 3D shape distribution into three key components: part stylizers, transformation samplers, and a cross diffusion network.

We present Adaptive IMLE, a generative modeling approach that covers all the modes and produces high-quality results. Adaptive IMLE is capble of learning from a few samples from scratch without any auxiliary datasets. We apply our method to the challenging task of few-shot unconditional image generation with as few as 100 data examples.

We present SCADE, a novel technique for NeRF reconstruction under sparse, unconstrained views for in-the-wild indoor scenes. We leverage on generalizable monocular depth priors and address to represent the inherent ambiguities of monocular depth by exploiting our ambiguity-aware depth estimates (leftmost). Our approach accounts for multimodality of both distributions using our novel space carving loss that seeks to disambiguate and find the common mode to fuse the information between different views.

We present Conditional Hierarchical IMLE (CHIMLE), a general purposed conditional image synthesis method that generates diverse and high-quality output images. We show CHIMLE significantly outperforms the prior best IMLE, GAN and diffusion-based methods in terms of image fidelity and mode coverage across four tasks, namely night-to-day, 16x single image super-resolution, image colourization and image decompression.