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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Training a 3D scene understanding model requires complicated humanannotations, which are laborious to collect and result in a model only encodingclose-set object semantics. In contrast, vision-language pre-training models(e.g., CLIP) have shown remarkable open-world reasoning properties. To thisend, we propose directly transferring CLIP's feature space to 3D sceneunderstanding model without any form of supervision. We first modify CLIP'sinput and forwarding process so that it can be adapted to extract dense pixelfeatures for 3D scene contents. We then project multi-view image features tothe point cloud and train a 3D scene understanding model with featuredistillation. Without any annotations or additional training, our modelachieves promising annotation-free semantic segmentation results onopen-vocabulary semantics and long-tailed concepts. Besides, serving as across-modal pre-training framework, our method can be used to improve dataefficiency during fine-tuning. Our model outperforms previous SOTA methods invarious zero-shot and data-efficient learning benchmarks. Most importantly, ourmodel successfully inherits CLIP's rich-structured knowledge, allowing 3D sceneunderstanding models to recognize not only object concepts but also open-worldsemantics.

CLIP-FO3D: Learning Free Open-world 3D Scene Representations from 2D Dense CLIP