LIPT: Latency-aware Image Processing Transformer

arXiv:2404.06075v1 Announce Type: new
Abstract: Transformer is leading a trend in the field of image processing. Despite the great success that existing lightweight image processing transformers have achieved, they are tailored to FLOPs or parameters reduction, rather than practical inference acceleration. In this paper, we present a latency-aware image processing transformer, termed LIPT. We devise the low-latency proportion LIPT block that substitutes memory-intensive operators with the combination of self-attention and convolutions to achieve practical speedup. Specifically, we propose a novel …

abstract arxiv cs.cv image image processing inference latency paper parameters practical processing success transformer transformers trend type