AI Paper Atlas

Research map / Detection and segmentation / 3D detection and driving

FlashBEV: Fast and Memory-Efficient Exact BEV Transformation with IO-Awareness

Shunsuke Yokokawa, Hironori Kasahara

ECCV 2026 · Detection and segmentation › 3D detection and driving · approach: Multi-camera BEV detection · 0 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
replace tensorized execution with fused FlashBEV kernelTensorized Sampling-VT · VT operator reference config (X=Y=200, Z=8, C=128)forward peak memory (MB): 1971.94 -> 52.82Improved
fuse VT into one kernelTensorized Sampling-VT · VT operator reference config (X=Y=200, Z=8, C=128)forward latency (ms): 9.18 -> 1.77Improved
fuse VT into one kernelTensorized Sampling-VT · VT operator reference config (X=Y=200, Z=8, C=128)backward latency (ms): 17.90 -> 3.27Improved
fuse VT into one kernelTensorized Sampling-VT · VT operator reference config (X=Y=200, Z=8, C=128)backward peak memory (MB): 1174.86 -> 105.63Improved
use FlashBEV as VT module end to endSimpleBEV pipeline · E2E reference config (X=Y=200, Z=8, C=128)E2E peak memory (MB): 2211.29 -> 876.76Improved
use FlashBEV as VT module end to endSimpleBEV pipeline · E2E reference config (X=Y=200, Z=8, C=128)E2E latency (ms): 67.19 -> 59.91Improved
use FlashBEV instead of tensorized VTSimpleBEV · nuScenes BEV detectionIoU: 46.9 -> 46.8No clear effect
replace SimpleBEV height conv with height sum + Conv2d (ablation)SimpleBEV · nuScenes BEV detectionIoU: 46.9 -> 46.9No clear effect
increase height bins Z from 8 to 32 in TensorRT (ablation)FlashBEV plugin · TensorRT pluginpeak memory (MB): 38 -> 38No clear effect
increase height bins Z from 8 to 32 in TensorRT (ablation)Tensorized Sampling-VT · TensorRT pluginpeak memory (MB): 1294 -> 4994Hurt
apply FlashSCA execution to BEVFormer SCABEVFormerSCA · BEVFormer SCA TensorRT (P=32)peak memory (MB): 3708 -> 1168Improved
apply FlashSCA execution to BEVFormer SCABEVFormerSCA · BEVFormer SCA TensorRT (P=32)latency (ms): 29.27 -> 13.90Improved

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