AI Paper Atlas

Research map / 3D vision / NeRF and radiance fields

BeNeRF:Neural Radiance Fields from a Single Blurry Image and Event Stream

Wenpu Li, Pian Wan, Peng Wang, Jinghang Li, Yi Zhou, Peidong Liu

ECCV 2024 · 3D vision › NeRF and radiance fields · approach: Sparse-input NeRF regularization · 15 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 event stream supervision to the blurry-image NeRF (ablation)BeNeRF · Livingroom (synthetic)PSNR: 24.40 -> 37.11Improved
use cubic B-spline instead of linear interpolation for camera trajectory (ablation)BeNeRF · Tanabata (synthetic)PSNR: 27.42 -> 32.14Improved
use cubic B-spline instead of linear interpolation for camera trajectory (ablation)BeNeRF · Livingroom (synthetic)PSNR: 34.16 -> 37.11Improved
increase number of virtual sharp images from 7 to 19 (ablation)BeNeRF · Tanabata (synthetic)PSNR: 29.28 -> 32.14Improved
increase number of virtual sharp images from 19 to 23 (ablation)BeNeRF · Livingroom (synthetic)PSNR: 37.11 -> 37.11No clear effect
increase event accumulation length alpha from 0.05 to 0.25 (ablation)BeNeRF · Livingroom and Tanabata (synthetic)PSNR: Livingroom 36.66 -> 37.11 (peak 37.21 at 0.15); Tanabata 31.86 -> 32.21Mixed
add BARF-style coarse-to-fine training (ablation)BeNeRF · Livingroom (synthetic)PSNR: 37.11 -> 34.20Hurt
NeRF-based single-image method vs single-image deblurring networksBeNeRF · Synthetic single-image deblurring (5 scenes, average)PSNR: 25.85 -> 33.65Improved
jointly optimize NeRF and trajectory vs EDI event-based deblurringBeNeRF · Synthetic event-enhanced deblurring (5 scenes, average)PSNR: 29.34 -> 33.65Improved
single image plus short event stream vs multi-view E2NeRFBeNeRF · E2NeRF synthetic (6 scenes, average)PSNR: 29.77 vs 29.66No clear effect
single image plus short event stream vs multi-view E2NeRFBeNeRF · E2NeRF synthetic (6 scenes, average)LPIPS: .0725 -> .0588Improved
single image plus short event stream vs multi-view E2NeRFBeNeRF · E2NeRF synthetic (6 scenes, average)SSIM: .9600 vs .9273Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 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