ImageNet

	A	B	C	D	E	F	G	H	I	J	K
1	Rank	Method	Top 1 Accuracy	Top 5 Accuracy	Number of params	Date	Other dataset	Ensemble	Multi-Crop	Added by AI Index	Reports Multi-Crop

2	1	Noisy Student (EfficientNet-L2)	87.40%	98.20%	480M	2019-11-11	TRUE
3	2	FixResNeXt-101 32x48d	86.40%	98.00%	829M	2019-06-14	TRUE
4	3	ResNeXt-101 32x48d	85.40%	97.60%	829M	2018-05-02	TRUE
5	4	ResNeXt-101 32x32d	85.10%	97.50%	466M	2018-05-02	TRUE
6	5	EfficientNet-B7 (RandAugment)	85.00%	97.20%	66M	2019-09-30	FALSE
7	6	ResNeXt-101 32x16d (semi-weakly sup.)	84.80%			2019-05-02	TRUE
8	7	EfficientNet-B7	84.40%	97.10%	66M	2019-05-28	FALSE
9	8	GPIPE	84.40%	97.00%	557M	2018-11-16	FALSE
10	9	ResNeXt-101 32x8d (semi-weakly sup.)	84.30%			2019-05-02	TRUE
11	10	ResNeXt-101 32×16d	84.20%	97.20%	194M	2018-05-02	TRUE
12	11	EfficientNet-B6	84.00%	96.90%	43M	2019-05-28	FALSE
13	12	AmoebaNet-A	83.90%	96.60%	469M	2018-02-05	FALSE
14	13	FixPNASNet-5	83.70%	96.80%	86.1M	2019-06-14	FALSE
15	14	MultiGrain PNASNet (500px)	83.60%	96.70%	86M	2019-02-14	FALSE
16	15	ResNeXt-101 32x4d (semi-weakly sup.)	83.40%			2019-05-02	TRUE
17	16	EfficientNet-B5	83.30%	96.70%	30M	2019-05-28	FALSE
18	17	MultiGrain SENet154 (400px)	83.00%	96.50%	86M	2019-02-14	FALSE
19	18	Oct-ResNet-152 (SE)	82.90%	96.30%	67M	2019-04-10	FALSE
20	19	PNASNet-5	82.90%	96.20%	86.1M	2017-12-02	FALSE
21	20	NASNET-A(6)	82.70%	96.20%	88.9M	2017-07-21	FALSE
22	21	SENet-154	82.70%	96.20%	146M	2017-09-05	FALSE
23	22	EfficientNet-B4	82.60%	96.30%	19M	2019-05-28	FALSE
24	23	FixResNet-50 Billion	82.50%	96.60%	25.6M	2019-06-14	TRUE
25	24	SCARLET-A4	82.30%	96.00%	27.8M	2019-09-01	FALSE
26	25	ResNeXt-101 64x4 + multi-scale dense testing	82.30%	96.30%		2016-11-16	FALSE	FALSE	TRUE	TRUE
27	26	ResNeXt-101 32x8d	82.20%	96.40%	88M	2018-05-02	TRUE
28	27	Inception ResNet V2 + multi-crop, multi-scale	82.20%	96.30%		2016-02-23	FALSE	FALSE	TRUE	TRUE
29	28	DPN-131 (320x320, Mean-Max Pooling)	81.45%	95.80%	80M	2017-07-06	FALSE
30	29	DPN-131 (320x320)	81.38%	95.80%	80M	2017-07-06	FALSE
31	30	PolyNet	81.30%	95.80%	92M	2016-11-17	FALSE				TRUE
32	31	DPN-98 (320x320, Mean-Max Pooling)	81.28%	95.60%		2017-07-06	FALSE
33	32	ResNet-50 (semi-weakly sup.)	81.20%			2019-05-02	TRUE
34	33	EfficientNet-B3	81.10%	95.50%	12M	2019-05-28	FALSE
35	34	DPN-98 (320x320)	81.06%	95.60%		2017-07-06	FALSE
36	35	DPN-92 (320x320, Mean-Max Pooling)	80.96%	95.50%		2017-07-06	FALSE
37	36	ResNeXt-101 64x4	80.90%	95.60%	83.6M	2016-11-16	FALSE
38	37	DPN-92 (320x320)	80.66%	95.30%		2017-07-06	FALSE
39	38	Inception ResNet V2	80.10%	95.10%	55.8M	2016-02-23	FALSE
40	39	Inception ResNet V1	88.70%	94.50%		2016-02-23	FALSE		FALSE
41	40	Inception-v4	80.00%	95.00%		2016-02-23	FALSE		FALSE
42	41	RandWire-WS	80.10%	94.80%		2019-04-02	FALSE
43	42	DPN-131 (224x224)	80.07%	94.90%	80M	2017-07-06	FALSE
44	43	DPN-98 (224x224)	79.95%	94.90%		2017-07-06	FALSE
45	44	ScaleNet-152	79.94%	94.80%		2019-04-20	FALSE
46	45	ResNet-200	79.90%	95.20%		2016-03-16	FALSE
47	46	Modified Aligned Xception	79.81%	94.80%		2018-02-07	FALSE
48	47	SKNet-101	79.81%			2019-03-15	FALSE
49	48	FixResNet-50 CutMix	79.80%	94.90%	25.6M	2019-06-14	FALSE
50	49	EfficientNet-B2	79.80%	94.90%	9.2M	2019-05-28	FALSE
51	50	Res2Net-DLA-60	79.47%			2019-04-02	FALSE
52	51	MultiGrain R50-AA-500	79.40%	94.80%		2019-02-14	FALSE
53	52	DPN-92 (224x224)	79.27%	94.60%		2017-07-06	FALSE
54	53	DenseNet-264	79.20%	94.70%		2016-08-25	FALSE
55	54	JFT-300M Finetuning	79.20%	94.70%		2017-07-10	TRUE
56	55	ScaleNet-101	79.18%	94.60%		2019-04-20	FALSE
57	56	AA-ResNet-152	79.10%	94.60%		2019-04-22	FALSE
58	57	FixResNet-50	79.10%	94.60%	25.6M	2019-06-14	FALSE
59	58	Xception	79%	94.50%	22.8M	2016-10-07	FALSE
60	59	ResNet-152 + SWA	78.94%			2018-03-14	FALSE
61	60	ECA-Net (ResNet-152)	78.92%	94.60%	57.40M	2019-10-08	FALSE
62	61	MixNet-L	78.90%	94.20%	7.3M	2019-07-22	FALSE
63	62	Inception V3	78.80%	94.40%	23.8M	2015-12-02	FALSE				TRUE
64	63	EfficientNet-B1	78.80%	94.40%	7.8M	2019-05-28	FALSE
65	64	SGE-ResNet101	78.80%	94.40%		2019-05-23	FALSE
66	65	ECA-Net (ResNet-101)	78.65%	94.30%	42.49M	2019-10-08	FALSE
67	66	ResNet-152	78.57%	94.30%		2015-12-10	FALSE
68	67	DenseNet-201	78.54%	94.50%	20M	2016-08-25	FALSE
69	68	DPN-68 (320x320, Mean-Max Pooling)	78.49%	94.50%		2017-07-06	FALSE
70	69	SRM-ResNet-101	78.47%			2019-03-26	FALSE
71	70	DenseNet-161 + SWA	78.44%			2018-03-14	FALSE
72	71	ResNet-50 (CutMix)	78.40%			2019-05-13	FALSE
73	72	ResNet-101	78.25%	94.00%	40M	2015-12-10	FALSE
74	73	MultiGrain R50-AA-224	78.20%	93.90%		2019-02-14	FALSE
75	74	WRN-50-2-bottleneck	78.10%	94.00%	68.9M	2016-05-23	FALSE
76	75	DenseNet-169	77.92%	94.10%		2016-08-25	FALSE
77	76	DPN-68 (320x320)	77.85%	94.10%		2017-07-06	FALSE
78	77	ECA-Net (ResNet-50)	77.48%	93.70%	24.37M	2019-10-08	FALSE
79	78	ResNet-50-D	77.16%	93.50%	25M	2018-12-04	FALSE
80	79	ResNet-50	77.15%	93.30%		2015-12-10	FALSE
81	80	MixNet-M	77.00%	93.30%	5.0M	2019-07-22	FALSE
82	81	SCARLET-A	76.90%	93.40%	6.7M	2019-08-16	FALSE
83	82	MnasNet-A3	76.70%	93.30%	5.2M	2018-07-31	FALSE
84	83	DPN-68 (224x224)	76.43%	93.10%		2017-07-06	FALSE
85	84	DenseNet-121	76.39%	93.30%		2016-08-25	FALSE
86	85	EfficientNet-B0	76.30%	93.20%	5.3M	2019-05-28	FALSE
87	86	SCARLET-B	76.30%	93.00%	6.5M	2019-08-16	FALSE
88	87	MoGA-A	75.90%	92.80%	5.1M	2019-08-01	FALSE
89	88	DenseNAS-A	75.90%	92.60%		2019-06-23	FALSE
90	89	FractalNet-34	75.88%	92.60%		2016-05-24	FALSE
91	90	MixNet-S	75.80%	92.80%	4.1M	2019-07-22	FALSE
92	91	CoordConv ResNet-50	75.74%	92.80%		2018-07-09	FALSE
93	92	PReLU-Net	75.73%	92.60%		2015-02-06	FALSE
94	93	LR-Net-26	75.70%	92.60%		2019-04-25	FALSE
95	94	MnasNet-A2	75.60%	92.70%	4.8M	2018-07-31	FALSE
96	95	SCARLET-C	75.60%	92.60%	6.0M	2019-08-16	FALSE
97	96	ShuffleNet V2	75.40%			2018-07-30	FALSE
98	97	MnasNet-A1	75.20%	92.50%	3.9M	2018-07-31	FALSE
99	98	MobileNet V3-Large 1.0	75.20%		5.4M	2019-05-06	FALSE
100	99	MultiGrain NASNet-A-Mobile (350px)	75.10%	92.50%		2019-02-14	FALSE