AI Paper Atlas

Research map / 3D vision / 3D generation

PromptDepth: Efficient and Promptable Geometric 3D Vision Model for Embodied Intelligence

Xianyun Wang, Jiaxu Miao, Tian Xu, Siyuan Wang, Yuehao Li, Haoyang Hu, Jun Xiao, Yonghong Tian, Jun Yu

CVPR 2026 · 3D vision › 3D generation · approach: Semantic scene completion

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add Gram anchoring loss to joint depth and instance training (ablation)PromptDepth · KITTI depth, multi-taskAbs rel: 0.091 -> 0.075Improved
add Gram anchoring loss to joint training (ablation)PromptDepth · Interactive segmentation, multi-taskmIoU: 62.4 -> 67.1Improved
add ILDS loss to joint training (ablation)PromptDepth · KITTI depth, multi-taskAbs rel: 0.085 -> 0.075Improved
add ILDS loss to joint training (ablation)PromptDepth · Interactive segmentation, multi-taskmIoU: 66.8 -> 67.1Improved
joint multi-task training with ILDS and Gram anchoring vs single-task depth training (ablation)PromptDepth · KITTI depthAbs rel: 0.081 -> 0.075Improved
joint multi-task training vs single-task segmentation training (ablation)PromptDepth · Interactive segmentationmIoU: 67.0 -> 67.1No clear effect
naive joint training of depth and instance tasks without Gram anchoring (ablation)PromptDepth · KITTI depth vs single-taskdelta1: 0.942 -> 0.924Hurt
train on high-fidelity post-rendered MRQ synthetic data (ablation)PromptDepth · DAVIS-17 zero-shot trackingJ&F-Mean: 81.2 -> 83.3Improved
train on high-fidelity post-rendered MRQ synthetic data (ablation)PromptDepth · YouTube-VOS zero-shot trackingJ&F-Mean: 75.0 -> 76.8Improved
use unified promptable model instead of SAM+DAv2 compositeembodied 3D instance-aware geometry · RTX 4090 float32 latency, monototal latency (ms): 95.53 -> 39.09Improved
use unified promptable model instead of SAM2+VDA compositestereo tracking · RTX 4090 float32 latency, stereototal latency (ms): 188.46 -> 44.92Improved
train on instance-level synthetic data with unified model vs SAMinteractive segmentation by point · GraspNetmIoU: 0.6036 -> 0.7863Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy