notesum.ai

Published at November 26

MAT: Multi-Range Attention Transformer for Efficient Image Super-Resolution

cs.CV

Released Date: November 26, 2024

Authors: Chengxing Xie¹, Xiaoming Zhang¹, Kai Zhang², Linze Li¹, Yuqian Fu³, Biao Gong⁴, Tianrui Li¹

Aff.: ¹Southwest Jiaotong University; ²Nanjing University; ³ETH Zurich; ⁴Ant Group

Arxiv: http://arxiv.org/abs/2411.17214v1

Scale	Method	Annual	Params. (K)	Multi-Adds (G)	Set5 PSNR / SSIM	Set14 PSNR / SSIM	B100 PSNR / SSIM	Urban100 PSNR / SSIM	Manga109 PSNR / SSIM
$\times 2$	EDSR-baseline	CVPRW17	1370	316.3	37.99 / 0.9604	33.57 / 0.9175	32.16 / 0.8994	31.98 / 0.9272	38.54 / 0.9769
	IMDN	MM19	694	158.8	38.00 / 0.9605	33.63 / 0.9177	32.19 / 0.8996	32.17 / 0.9283	38.88 / 0.9774
	LatticeNet	ECCV20	756	169.5	38.06 / 0.9607	33.70 / 0.9187	32.20 / 0.8999	32.25 / 0.9288	-
	SwinIR-light	ICCVW21	910	252.9	38.14 / 0.9611	33.86 / 0.9206	32.31 / 0.9012	32.76 / 0.9340	39.12 / 0.9783
	SwinIR-NG	CVPR23	1181	274.1	38.17 / 0.9612	33.94 / 0.9205	32.31 / 0.9013	32.78 / 0.9340	39.20 / 0.9781
	Omni-SR	CVPR23	798	-	38.22 / 0.9613	33.98 / 0.9210	32.36 / 0.9020	33.05 / 0.9363	39.28 / 0.9784
	CRAFT	ICCV23	738	197.2	38.23 / 0.9615	33.92 / 0.9211	32.33 / 0.9016	32.86 / 0.9343	39.39 / 0.9786
	SRFormer-light	ICCV23	853	236.2	38.23 / 0.9613	33.94 / 0.9209	32.36 / 0.9019	32.91 / 0.9353	39.28 / 0.9785
	MambaIR-light	ECCV24	1363	278.9	38.16 / 0.9610	34.00 / 0.9212	32.34 / 0.9017	32.92 / 0.9356	39.31 / 0.9779
	MAT-light	Ours	694	189.6	38.28 / 0.9617	34.11 / 0.9227	32.41 / 0.9029	33.22 / 0.9381	39.46 / 0.9789
$\times 3$	EDSR-baseline	CVPRW17	1555	160.2	34.37 / 0.9270	30.28 / 0.8417	29.09 / 0.8052	28.15 / 0.8527	33.45 / 0.9439
	IMDN	MM19	703	71.5	34.36 / 0.9270	30.32 / 0.8417	29.09 / 0.8046	28.17 / 0.8519	33.61 / 0.9445
	LatticeNet	ECCV20	765	76.3	34.40 / 0.9272	30.32 / 0.8416	29.10 / 0.8049	28.19 / 0.8513	-
	SwinIR-light	ICCVW21	918	114.5	34.62 / 0.9289	30.54 / 0.8463	29.20 / 0.8082	28.66 / 0.8624	33.98 / 0.9478
	SwinIR-NG	CVPR23	1190	114.1	34.64 / 0.9293	30.58 / 0.8471	29.24 / 0.8090	28.75 / 0.8639	34.22 / 0.9488
	Omni-SR	CVPR23	807	-	34.70 / 0.9294	30.57 / 0.8469	29.28 / 0.8094	28.84 / 0.8656	34.22 / 0.9487
	CRAFT	ICCV23	744	87.5	34.71 / 0.9295	30.61 / 0.8469	29.24 / 0.8093	28.77 / 0.8635	34.29 / 0.9491
	SRFormer-light	ICCV23	861	104.8	34.67 / 0.9296	30.57 / 0.8469	29.26 / 0.8099	28.81 / 0.8655	34.19 / 0.9489
	MambaIR-light	ECCV24	1371	124.6	34.72 / 0.9296	30.63 / 0.8475	29.29 / 0.8099	29.00 / 0.8689	34.39 / 0.9495
	MAT-light	Ours	703	85.0	34.79 / 0.9303	30.68 / 0.8491	29.32 / 0.8116	29.03 / 0.8698	34.49 / 0.9505
$\times 4$	EDSR-baseline	CVPRW17	1518	114.0	32.09 / 0.8938	28.58 / 0.7813	27.57 / 0.7357	26.04 / 0.7849	30.35 / 0.9067
	IMDN	MM19	715	40.9	32.21 / 0.8948	28.58 / 0.7811	27.56 / 0.7353	26.04 / 0.7838	30.45 / 0.9075
	LatticeNet	ECCV20	777	43.6	32.18 / 0.8943	28.61 / 0.7812	27.57 / 0.7355	26.14 / 0.7844	-
	SwinIR-light	ICCVW21	930	65.2	32.44 / 0.8976	28.77 / 0.7858	27.69 / 0.7406	26.47 / 0.7980	30.92 / 0.9151
	SwinIR-NG	CVPR23	1201	63.0	32.44 / 0.8980	28.83 / 0.7870	27.73 / 0.7418	26.61 / 0.8010	31.09 / 0.9161
	Omni-SR	CVPR23	819	-	32.49 / 0.8988	28.78 / 0.7859	27.71 / 0.7415	26.64 / 0.8018	31.02 / 0.9151
	CRAFT	ICCV23	753	52.4	32.52 / 0.8989	28.85 / 0.7872	27.72 / 0.7418	26.56 / 0.7995	31.18 / 0.9168
	SRFormer-light	ICCV23	873	62.8	32.51 / 0.8988	28.82 / 0.7872	27.73 / 0.7422	26.67 / 0.8032	31.17 / 0.9165
	MambaIR-light	ECCV24	1383	70.8	32.51 / 0.8993	28.85 / 0.7876	27.75 / 0.7423	26.75 / 0.8051	31.26 / 0.9175
	MAT-light	Ours	714	48.5	32.61 / 0.8998	28.92 / 0.7897	27.79 / 0.7444	26.83 / 0.8088	31.38 / 0.9192