AI Paper Atlas

Research map / Point clouds / Point cloud classification

Take-A-Photo: 3D-to-2D Generative Pre-training of Point Cloud Models

Ziyi Wang, Xumin Yu, Yongming Rao, Jie Zhou, Jiwen Lu

ICCV 2023 · Point clouds › Point cloud classification · approach: Self-supervised point cloud pretraining · 36 citations (OpenAlex)

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 TAP 3D-to-2D pre-trainingPointMLP · ScanObjectNN PB_T50_RSaccuracy: 87.4 -> 88.5Improved
add TAP pre-trainingDGCNN · ScanObjectNN PB_T50_RSaccuracy: 86.1 -> 86.6Improved
add TAP pre-trainingPointMLP · ModelNet40accuracy: 93.7 -> 94.0Improved
add TAP pre-trainingPointMLP · ShapeNetPartmIoU_I: 86.1 -> 86.9Improved
pre-train encoder with TAP3DETR · ScanNetV2 detectionAP0.5: 37.9 -> 41.4Improved
use cross-attention that learns projection instead of direct projection (ablation)TAP photograph module · ScanObjectNN PB_T50_RSaccuracy: 87.6 -> 88.5Improved
add learnable memory pad token (ablation)TAP photograph module · ScanObjectNN PB_T50_RSaccuracy: 88.3 -> 88.5Improved
use formula-derived optical-line queries instead of learnable queries (ablation)TAP query generator · ScanObjectNN PB_T50_RSaccuracy: 87.5 -> 88.5Improved
include origin point coordinates in query (ablation)TAP query generator · ScanObjectNN PB_T50_RSaccuracy: 86.5 -> 88.5Improved
use 1024-dim cross-attention channels (no bottleneck) (ablation)TAP photograph module · ScanObjectNN PB_T50_RSaccuracy: 88.5 -> 85.3Hurt
add VGG perceptual loss to pixel loss (ablation)TAP objective · ScanObjectNN PB_T50_RSaccuracy: 88.5 -> 87.8Hurt
use foreground:background loss weight 20:1 (ablation)TAP objective · ScanObjectNN PB_T50_RSaccuracy: 87.1 -> 88.5Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 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