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

Feature representation learning is the key recipe for learning-basedMulti-View Stereo (MVS). As the common feature extractor of learning-based MVS,vanilla Feature Pyramid Networks (FPNs) suffer from discouraged featurerepresentations for reflection and texture-less areas, which limits thegeneralization of MVS. Even FPNs worked with pre-trained Convolutional NeuralNetworks (CNNs) fail to tackle these issues. On the other hand, VisionTransformers (ViTs) have achieved prominent success in many 2D vision tasks.Thus we ask whether ViTs can facilitate feature learning in MVS? In this paper,we propose a pre-trained ViT enhanced MVS network called MVSFormer, which canlearn more reliable feature representations benefited by informative priorsfrom ViT. The finetuned MVSFormer with hierarchical ViTs of efficient attentionmechanisms can achieve prominent improvement based on FPNs. Besides, thealternative MVSFormer with frozen ViT weights is further proposed. This largelyalleviates the training cost with competitive performance strengthened by theattention map from the self-distillation pre-training. MVSFormer can begeneralized to various input resolutions with efficient multi-scale trainingstrengthened by gradient accumulation. Moreover, we discuss the merits anddrawbacks of classification and regression-based MVS methods, and furtherpropose to unify them with a temperature-based strategy. MVSFormer achievesstate-of-the-art performance on the DTU dataset. Particularly, MVSFormer ranksas Top-1 on both intermediate and advanced sets of the highly competitiveTanks-and-Temples leaderboard.

MVSFormer: Multi-View Stereo by Learning Robust Image Features and Temperature-based Depth