English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

The task of novel view synthesis aims to generate unseen perspectives of anobject or scene from a limited set of input images. Nevertheless, synthesizingnovel views from a single image still remains a significant challenge in therealm of computer vision. Previous approaches tackle this problem by adoptingmesh prediction, multi-plain image construction, or more advanced techniquessuch as neural radiance fields. Recently, a pre-trained diffusion model that isspecifically designed for 2D image synthesis has demonstrated its capability inproducing photorealistic novel views, if sufficiently optimized on a 3Dfinetuning task. Although the fidelity and generalizability are greatlyimproved, training such a powerful diffusion model requires a vast volume oftraining data and model parameters, resulting in a notoriously long time andhigh computational costs. To tackle this issue, we propose Efficient-3DiM, asimple but effective framework to learn a single-image novel-view synthesizer.Motivated by our in-depth analysis of the inference process of diffusionmodels, we propose several pragmatic strategies to reduce the training overheadto a manageable scale, including a crafted timestep sampling strategy, asuperior 3D feature extractor, and an enhanced training scheme. When combined,our framework is able to reduce the total training time from 10 days to lessthan 1 day, significantly accelerating the training process under the samecomputational platform (one instance with 8 Nvidia A100 GPUs). Comprehensiveexperiments are conducted to demonstrate the efficiency and generalizability ofour proposed method.

Efficient-3DiM: Learning a Generalizable Single-image Novel-view Synthesizer in One Day