AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

Efficient Visual Computing With Camera RAW Snapshots

Zhihao Li, Ming Lu, Xu Zhang, Xin Feng, M. Salman Asif, Zhan Ma

TPAMI 2024 · Vision-language and multimodal › Human and embodied modeling · approach: Human-like visual cognition in VLMs · 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
use the same ISP for training and testing instead of mismatched ISPs (ablation)RGB-Vision YOLOv8-S · MultiRAW detectionmAP: 0.397 -> 0.529Improved
use expert-calibrated iPhone ISP instead of tuned AX620A ISP (ablation)RGB-Vision YOLOv8-S · MultiRAW detectionmAP: 0.529 -> 0.546Improved
tune the AX620A ISP instead of default settings (ablation)RGB-Vision YOLOv8-S · MultiRAW detectionmAP: 0.173 -> 0.397Improved
replace ISP + RGB detector with RAW-domain detectorYOLOv8-S · MultiRAW detectionmAP: 0.546 -> 0.546No clear effect
replace ISP + RGB detector with RAW-domain detectorYOLOv8-S · AX620A hardwarelatency: 72% latency reductionImproved
replace ISP + RGB classifier with RAW-domain classifierMobileNet-V1 · Low-light classificationTop-1 accuracy: 0.0 -> 19.8Improved
replace ISP + RGB classifier with RAW-domain classifierMobileNet-V1 · Low-light classificationenergy: 0.128 J -> 0.006 JImproved
use AX620A AI-ISP instead of standard ISP (ablation)RGB-Vision MobileNet-V1 · Low-light classificationTop-1 accuracy: 0.0 -> 0.0No clear effect
replace ISP + RGB detector with RAW-domain detectorYOLOv8-S · HDR detectionmAP: 56.8 -> 63.3Improved
use Unpaired CycleR2R simRAW instead of Mosaic RAWMobileNet-V1 · Low-light classificationTop-5 accuracy: 52.5 -> 72.1Improved
use Unpaired CycleR2R simRAW instead of CycleISPHRNetv2 · Segmentation on iPhone RAWmIoU: 43.6 -> 50.6Improved
train on simRAW instead of feeding real RAW to RGB modelHRNetv2 · Segmentation on iPhone RAWmIoU: 11.1 -> 50.6Improved

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