AI Paper Atlas

Research map / 3D vision / 3D generation

Pathformer3D: A 3D Scanpath Transformer for 360° Images

Rong Quan, yantao Lai, Mengyu Qiu, Dong Liang

ECCV 2024 · 3D vision › 3D generation · approach: Learning 3D reconstruction from images

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
replace SphereNet features with ViT patch features (ablation)Pathformer3D · Salient360!DTW: 1515.883 -> 1739.714Hurt
add 3D Transformer encoder layers (ablation)Pathformer3D · Salient360!DTW: 1798.340 -> 1515.883Improved
use 3D mixture density network instead of linear regression with MSE loss (ablation)Pathformer3D · Salient360!SS: 0.095 -> 0.302Improved
use K=5 Gaussian kernels instead of K=1, 3 or 8 (ablation)Pathformer3D · Salient360!DTW: K=3: 2073.905, K=8: 1700.877 -> K=5: 1515.883Mixed
autoregressive decoder instead of parallel decoder (ablation)Pathformer3D · Salient360!TDE: 23.726 -> 19.261Improved
predict scanpaths in 3D spherical coordinates instead of 2D equirectangular planePathformer3D · Salient360!DTW: 1706.661 -> 1515.883Improved
autoregressive decoding plus mixture density network adds inference costPathformer3D · efficiency comparisonrunning time (ms): 118.250 -> 266.420Hurt

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