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

In this work, we propose a novel single-shot and keypoints-based frameworkfor monocular 3D objects detection using only RGB images, called KM3D-Net. Wedesign a fully convolutional model to predict object keypoints, dimension, andorientation, and then combine these estimations with perspective geometryconstraints to compute position attribute. Further, we reformulate thegeometric constraints as a differentiable version and embed it into the networkto reduce running time while maintaining the consistency of model outputs in anend-to-end fashion. Benefiting from this simple structure, we then propose aneffective semi-supervised training strategy for the setting where labeledtraining data is scarce. In this strategy, we enforce a consensus prediction oftwo shared-weights KM3D-Net for the same unlabeled image under different inputaugmentation conditions and network regularization. In particular, we unify thecoordinate-dependent augmentations as the affine transformation for thedifferential recovering position of objects and propose a keypoints-dropoutmodule for the network regularization. Our model only requires RGB imageswithout synthetic data, instance segmentation, CAD model, or depth generator.Nevertheless, extensive experiments on the popular KITTI 3D detection datasetindicate that the KM3D-Net surpasses all previous state-of-the-art methods inboth efficiency and accuracy by a large margin. And also, to the best of ourknowledge, this is the first time that semi-supervised learning is applied inmonocular 3D objects detection. We even surpass most of the previous fullysupervised methods with only 13\% labeled data on KITTI.

Monocular 3D Detection with Geometric Constraints Embedding and Semi-supervised Training