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

Transformer, as an alternative to CNN, has been proven effective in manymodalities (e.g., texts and images). For 3D point cloud transformers, existingefforts focus primarily on pushing their accuracy to the state-of-the-artlevel. However, their latency lags behind sparse convolution-based models (3xslower), hindering their usage in resource-constrained, latency-sensitiveapplications (such as autonomous driving). This inefficiency comes from pointclouds' sparse and irregular nature, whereas transformers are designed fordense, regular workloads. This paper presents FlatFormer to close this latencygap by trading spatial proximity for better computational regularity. We firstflatten the point cloud with window-based sorting and partition points intogroups of equal sizes rather than windows of equal shapes. This effectivelyavoids expensive structuring and padding overheads. We then applyself-attention within groups to extract local features, alternate sorting axisto gather features from different directions, and shift windows to exchangefeatures across groups. FlatFormer delivers state-of-the-art accuracy on WaymoOpen Dataset with 4.6x speedup over (transformer-based) SST and 1.4x speedupover (sparse convolutional) CenterPoint. This is the first point cloudtransformer that achieves real-time performance on edge GPUs and is faster thansparse convolutional methods while achieving on-par or even superior accuracyon large-scale benchmarks.

FlatFormer: Flattened Window Attention for Efficient Point Cloud Transformer