Sample stimuli

How to use

from brainscore_vision import load_benchmark
benchmark = load_benchmark("Geirhos2021eidolonI-error_consistency")
score = benchmark(my_model)

Benchmark API

Code examples

Model scores

Score Legend Min Alignment Max Alignment

Rank	Model	Score
1	vision_transformer_vit_large_patch16_224	.789
2	effnetb1_cutmixpatch_SAM_robust32_avge6e8e9e10_manylayers_324x288	.783
3	effnetb1_cutmixrespatch_SAM_robust32_e10_manylayers_324x288	.777
4	effnetb1_cutmixrespatch_SAM_robust32_e2_manylayers_324x288	.776
5	resnext101_32x16d_wsl	.731
6	vit_base_patch16_clip_224:openai_ft_in1k	.724
7	SWSL_resnext101_32x8d	.715
8	resnet50_finetune_cutmix_e3_robust_linf8255_e0_247x234	.698
9	vit_large_patch14_clip_224:openai_ft_in12k_in1k	.689
10	ViT_B_16_imagenet1k	.677
11	resnext101_32x48d_wsl	.667
12	resnet_101_v2	.663
13	effnetb1_cutmixpatch_augmix_robust32_avge4e7_manylayers_324x288	.661
14	resnext101_32x8d_wsl	.655
15	CLIP_ViT-B_32	.651
16	SWSL_resnet50	.647
17	AdvProp_efficientnet-b7	.646
18	resnet_152_v2	.638
19	ViT_L_32_imagenet1k	.637
20	resnext101_32x32d_wsl	.637
21	vit_large_patch14_clip_224:laion2b_ft_in1k	.636
22	effnetb1_cutmixpatch_augmix_robust32_e5_324x288	.629
23	resnet101	.629
24	effnetv2m_custom384	.621
25	AT_efficientnet-b4	.619
26	AdvProp_efficientnet-b4	.619
27	vit_large_patch14_clip_224:openai_ft_in1k	.619
28	vit_large_patch14_clip_224:laion2b_ft_in12k_in1k	.615
29	effnetb1_272x240	.614
30	vit_relpos_base_patch16_clsgap_224:sw_in1k	.610
31	resnet152_imagenet_full	.607
32	focalnet_tiny_lrf_in1k	.603
33	cvt_cvt-13-384-in22k_finetuned-in1k_4_LucyV4	.593
34	pnasnet_large	.593
35	pnasnet_large-tf	.584
36	efficientnet_b0	.583
37	fixres_resnext101_32x48d_wsl	.582
38	convnext_large_mlp:clip_laion2b_augreg_ft_in1k_384	.577
39	vit_base_patch16_clip_224:openai_ft_in12k_in1k	.575
40	vit_large_patch14_clip_336:laion2b_ft_in1k	.574
41	convnext_xlarge:fb_in22k_ft_in1k	.574
42	vit_huge_patch14_clip_336:laion2b_ft_in12k_in1k	.567
43	cvt_cvt-21-384-in22k_finetuned-in1k_4_LucyV4	.556
44	AdvProp_efficientnet-b6	.553
45	cvt_cvt-13-384-in1k_4_LucyV4	.551
46	vit_large_patch14_clip_336:openai_ft_in12k_in1k	.549
47	resnet_101_v1	.546
48	AdvProp_efficientnet-b2	.541
49	AT_efficientnet-b2	.541
50	cvt_cvt-13-224-in1k_4_LucyV4	.537
51	vit_relpos_base_patch32_plus_rpn_256:sw_in1k	.529
52	vonegrcnn_62e_nobn	.527
53	cvt_cvt-w24-384-in22k_finetuned-in1k_4	.527
54	antialias-resnet152	.524
55	ViT_B_32_imagenet1k	.522
56	resnet-152_v2-tf	.520
57	inception_v4	.519
58	CLIP_resnet50	.513
59	CLIP_resnet50_float32	.513
60	custom_model_cv_18_dagger_408	.512
61	nasnet_large	.505
62	nasnet_large-tf	.499
63	convnext_tiny_imagenet_full_seed-0	.498
64	vonegrcnn_47e	.497
65	convnext_large:fb_in22k_ft_in1k	.492
66	resnet_50_v2	.491
67	vonegrcnn_52e_full	.487
68	convnext_femto_ols:d1_in1k	.482
69	convnext_xxlarge:clip_laion2b_soup_ft_in1k	.481
70	vonegrcnn_62e_nobn	.476
71	convnext_base:clip_laiona_augreg_ft_in1k_384	.472
72	convnext_tiny:in12k_ft_in1k	.471
73	resnet_152_v1	.470
74	cvt_cvt-21-384-in1k_4_LucyV4	.464
75	resnet50_linf_4_robust	.459
76	resnet-152_v2_pytorch	.459
77	resnet-152_v1-tf	.459
78	vonegrcnn_47e	.456
79	efficientnet-b7	.453
80	densenet-201-keras	.448
81	densenet_201_pytorch	.446
82	inception_v3_pytorch	.445
83	inception_v4-tf	.443
84	resnet50_imagenet_full	.443
85	densenet-201	.443
86	resnet34_imagenet_full	.442
87	antialiased-rnext101_32x8d	.436
88	vonegrcnn_52e_full	.435
89	resnet50-SIN_IN_IN	.432
90	resnet_SIN_IN_FT_IN	.432
91	vit_huge_patch14_clip_224:laion2b_ft_in12k_in1k	.431
92	resnet50-VITO-8deg-cc	.430
93	cvt_cvt-21-224-in1k_4_LucyV4	.429
94	vit_tiny_r_s16_p8_384:augreg_in21k_ft_in1k	.424
95	voneresnet-50-non_stochastic	.417
96	convnext_small_imagenet_100_seed-0	.417
97	deit_large_imagenet_full_seed-0	.416
98	resnet101_imagenet_full	.412
99	CLIP-RN50	.409
100	BiT-S-R50x3	.408
101	convnext_base_imagenet_full_seed-0	.406
102	voneresnet-50-robust	.403
103	xception-keras	.401
104	resnet-50-pytorch	.400
105	resnet50-sup	.400
106	resnet50_tutorial	.400
107	resnet_50_v1	.400
108	resnet50-moclr8deg	.394
109	resnet50_random_linf8_perturb	.393
110	ReAlnet03	.384
111	efficientnet_b2_imagenet_full	.383
112	resnet-50_v2-tf	.378
113	inception_v3-tf	.377
114	resnet50_random_l2_perturb	.371
115	convnext_small_imagenet_full_seed-0	.367
116	resnet-101_v1-tf	.363
117	resnet-50_v1-tf	.359
118	ReAlnet02	.355
119	resnet50-SIN_IN	.355
120	convnext_large_imagenet_full_seed-0	.354
121	Res2Net50_26w_4s	.350
122	AdvProp_efficientnet-b8	.349
123	densenet-121-keras	.346
124	resnet50	.344
125	inception_v3	.341
126	deit_small_imagenet_full_seed-0	.338
127	BiT-S-R152x2	.336
128	tv_efficientnet-b1	.336
129	nasnet_mobile	.335
130	ReAlnet04	.334
131	grcnn_109	.333
132	blt_vs	.331
133	ReAlnet07	.329
134	ReAlnet_incorrect	.328
135	ReAlnet01	.328
136	xception	.327
137	ReAlnet06	.326
138	grcnn	.325
139	inception_v1	.324
140	ReAlnet08	.324
141	mobilenet_v2_1.0_192-tf	.318
142	antialiased-r50	.318
143	ReAlnet05	.317
144	inception_v2-tf	.309
145	densenet-121	.304
146	mobilenet_v2_1_0_192	.301
147	deit_base_imagenet_full_seed-0	.299
148	vonealexnet_gaussian_noise_std4_fixed	.297
149	ReAlnet10	.297
150	CORnet-S	.297
151	effnetb1_cutmix_augmix_sam_e1_5avg_424x377	.296
152	resnet18_imagenet_full	.295
153	vgg_19	.295
154	cornet_s	.293
155	resnet50_robust_l2_eps1	.291
156	alexnet	.290
157	regnet_y_400mf	.288
158	mobilenet_v1_0.75_224	.287
159	resnet50_linf_8_robust	.287
160	texture_shape_resnet50_trained_on_SIN	.286
161	swin_small_patch4_window7_224:ms_in22k_ft_in1k	.281
162	voneresnet-50	.281
163	alexnet	.281
164	alexnet	.281
165	alexnet	.281
166	alexnet	.281
167	alexnet	.281
168	alexnet	.281
169	alexnet	.281
170	alexnet	.281
171	alexnet-baseline	.281
172	alexnet	.281
173	ReAlnet09	.277
174	mobilenet_v2_1_0_224	.276
175	densenet-169-keras	.274
176	resnet50_l2_3_robust	.272
177	mobilenet_v1_1.0_192	.269
178	konkle_alexnetgn_ipcl_ref12_supervised_ipcl_aug	.264
179	alexnet_random_linf8_perturb	.260
180	resnet50_moco_v2	.256
181	vgg_16	.255
182	mobilenet_v2_0_5_224	.253
183	alexnet_reduced_aliasing_early_checkpoint	.252
184	resnet50_byol	.251
185	alexnet_l2_3_robust	.250
186	efficientnet_b1_imagenet_full	.249
187	mobilenet_v2_1_3_224	.247
188	resnet-18	.245
189	resnet50-SIN	.243
190	imagenet_l2_3_0	.242
191	resnet-50-robust	.242
192	resnet50_robust_l2_eps3	.242
193	densenet-169	.241
194	mobilenet_v2_1_0_128	.239
195	BiT-S-R152x4	.238
196	mobilenet_v2_1_0_160	.235
197	mobilenet_v1_1.0_224	.230
198	mobilenet_v2_1.0_160-tf	.229
199	inception_v1-tf	.223
200	cv_18_dagger_408_pretrained	.223
201	shufflenet_v2_x1_0	.222
202	alexnet_linf_8_robust	.218
203	efficientnet_b0_imagenet_full	.217
204	BiT-S-R101x3	.213
205	resnet50-vitoimagevidnet8	.210
206	alexnet_random_l2_3_perturb	.210
207	mobilenet_v2_1.4_224-tf	.208
208	BiT-S-R50x1	.208
209	resnet-34	.208
210	mobilenet_v2_1.3_224-tf	.206
211	mobilenet_v2_0_75_160	.204
212	mobilenet_v1_1.0_160	.198
213	vgg-16-keras	.197
214	mobilenet_v2_0.5_192-tf	.196
215	resnet50_simclr	.196
216	mobilenet_v2_1.0_224-tf	.194
217	mobilenet_v1_1.0_128	.192
218	BiT-S-R101x1	.188
219	squeezenet1_1	.182
220	mobilenet_v2_0.75_96	.181
221	mobilenet_v2_0_5_192	.180
222	alexnet_training_seed_05	.180
223	alexnet_training_seed_01	.180
224	alexnet_robust_correct	.179
225	vgg-19-keras	.179
226	deit_small_imagenet_100_seed-0	.171
227	mobilenet_v2_0.75_128	.169
228	konkle_alexnetgn_ipcl_ref01_primary_model	.168
229	mobilenet_v2_0.75_160-tf	.164
230	convnext_tiny_sup	.161
231	alexnet_early_checkpoint	.161
232	mobilenet_v1_0.25_192	.160
233	0.5x_resnet-18_LC_w_sh_1_iter	.160
234	mobilenet_v2_1-4_224_pytorch	.159
235	mobilenet_v2_1_4_224	.159
236	mobilenet_v2_0.5_224-tf	.156
237	artResNet18_1	.156
238	mobilenet_v2_0.35_160	.150
239	mobilenet_v1_0.75_128	.148
240	mobilenet_v2_1.0_128-tf	.143
241	mobilenet_v1_0.75_192	.141
242	resnet50_imagenet_100_seed-0	.140
243	AdvProp_efficientnet-b0	.140
244	mobilenet_v1_0.5_128	.138
245	imagenet_l2_10_0	.132
246	nasnet_mobile-tf	.132
247	mobilenet_v2_1.0_96	.131
248	mobilenet_v2_0.35_96	.131
249	texture_shape_alexnet_trained_on_SIN	.129
250	squeezenet1_0	.126
251	mobilenet_v1_0.5_192	.126
252	mobilenet_v1_0.25_160	.125
253	AlexNet_SIN	.122
254	mobilenet_v2_0.5_160	.120
255	mobilenet_v2_0_75_192	.120
256	mobilenet_v1_0.25_224	.113
257	mobilenet_v2_0.5_128	.113
258	mobilenet_v2_0.75_192-tf	.111
259	mobilenet_v1_0.75_160	.107
260	mobilenet_v1_0.5_224	.097
261	mobilenet_v2_0_75_224	.097
262	resnet50-meshes-lt-100-original-scratch	.094
263	deit_small_imagenet_1_seed-0	.094
264	deit_small_imagenet_10_seed-0	.094
265	resnet50_imagenet_10_seed-0	.093
266	alexnet_training_seed_06	.089
267	alexnet_training_seed_10	.085
268	alexnet_training_seed_02	.085
269	alexnet_training_seed_08	.085
270	alexnet_training_seed_04	.085
271	convnext_small_imagenet_10_seed-0	.084
272	alexnet_training_seed_07	.084
273	mobilenet_v2_0.5_96	.077
274	mobilenet_v2_0.35_128	.070
275	barlow-twins-resnet50	.063
276	resnet50-vicregl0p9	.059
277	mobilenet_v2_0.75_224-tf	.058
278	resnet50-meshes-lt-100-original-pretrained	.058
279	resnet50_FractalDB	.057
280	mobilenet_v2_0.35_224	.056
281	tf_efficientnetv2_s_in21ft1k_robust_linf12255_400x400	.055
282	resnet50-vicregl0p75	.051
283	resnet50-vicreg	.044
284	alexnet_training_seed_09	.042
285	resnet50-barlow	.041
286	mobilenet_v1_0.5_160	.038
287	mobilenet_v2_0.35_192	.036
288	my-model	.029
289	convnext_small_imagenet_1_seed-0	.015
290	mobilenet_v1_0.25_128	.013
291	resnet50-simclr-vissl	.009
292	mobilevit_small	.008
293	resnet50_imagenet_1_seed-0	.008
294	alexnet_training_seed_03	.002
295	resnet-50_untrained	.002
296	yudixie_resnet50_distance_rotation_0_240908	X
297	yudixie_resnet50_rotation_reg_0_240908	X
298	yudixie_resnet50_distance_translation_0_240908	X
299	resnet152_ecoset_full	X
300	yudixie_resnet18_distance_translation_0_240719	X
301	ReAlnet04_cornet	X
302	yudixie_resnet18_translation_rotation_0_240719	X
303	yudixie_resnet18_distance_rotation_0_240719	X
304	yudixie_resnet18_category_class_0_240719	X
305	ReAlnet05_cornet	X
306	yudixie_resnet18_random_0_240719	X
307	BiT-M-R50x1	X
308	yudixie_resnet18_cat_obj_class_all_latents_0_240719	X
309	resnet-18_test_m	X
310	yudixie_resnet18_distance_translation_rotation_0_240719	X
311	yudixie_resnet18_object_class_0_240719	X
312	yudixie_resnet18_imagenet1kpret_0_240719	X
313	ReAlnet06_cornet	X
314	yudixie_resnet50_translation_reg_0_240908	X
315	BiT-M-R50x3	X
316	0.5x_resnet-18_LC_w_sh_10_iter	X
317	yudixie_resnet50_translation_rotation_0_240908	X
318	ReAlnet07_cornet	X
319	resnet18-colorization	X
320	ReAlnet08_cornet	X
321	vggface	X
322	omnivore_swinB	X
323	ReAlnet09_cornet	X
324	my-model	X
325	FrankRobWobv0	X
326	omnivore_swinS	X
327	resnet18-instance_recognition	X
328	ViT-B/32	X
329	CORnet-Z	X
330	ReAlnet10_cornet	X
331	yudixie_resnet50_distance_translation_rotation_0_240908	X
332	r3m_resnet18	X
333	resnet18-deepcluster	X
334	CORnetZ_CIFAR10	X
335	resnet-18-LC_w_sh_100_iter_conv_init_m	X
336	0.5x_resnet-18_LC	X
337	fulltest_microblockvf_nobottleneck_freeat_nopretrain_eps2_m4	X
338	resnet18-local_aggregation	X
339	resnet50_ImageNet	X
340	resnet50_pretrained_with_retinal_waves	X
341	resnet50-cifar	X
342	resnet-18-LC_w_sh_100_iter_m	X
343	yudixie_resnet50_random_0_240908	X
344	alexnet_ks_torevert	X
345	openclip	X
346	resnet50	X
347	yudixie_resnet50_imagenet1kpret_0_240908	X
348	yudixie_resnet50_imagenet1kpret_0_240312	X
349	resnet-18-LC_w_sh_10_iter_m	X
350	yudixie_resnet50_cat_obj_class_all_latents_0_240908	X
351	yudixie_resnet50_object_class_0_240908	X
352	dorinet_cornet_z	X
353	resnet-18-LC_w_sh_1_iter_m	X
354	pixels	X
355	CORnetZ_CIFAR10_bs32_20_04	X
356	unet_entire	X
357	v1-pyr-nodown	X
358	dcgan	X
359	resnet-50x2_untrained	X
360	resnet-18_untrained	X
361	resnet-18-LC_untrained	X
362	omnivore_swinT	X
363	resnet18-contrastive_predictive	X
364	r3m_resnet34_nocrop	X
365	ViT_B_16	X
366	yudixie_resnet50_category_class_0_240908	X
367	resnet-18-LC_w_sh_10_iter_conv_init_m	X
368	resnet-18-LC_w_sh_1_iter_conv_init_m	X
369	resnet18-depth_prediction	X
370	ViT_B_32	X
371	ViT_L_32	X
372	resnet18-simclr	X
373	resnet18_imagenet21kP	X
374	resnet18-autoencoder	X
375	resnet18-relative_position	X
376	yudixie_resnet50_distance_reg_0_240908	X
377	ConvLSTM	X
378	PredRNN	X
379	SimVP	X
380	TAU	X
381	MIM	X
382	hmax	X
383	BiT-M-R152x2	X
384	RN50	X
385	0.5x_resnet-18_LC_w_sh_100_iter	X
386	pnasnet_large_pytorch	X
387	ReAlnet01_cornet	X
388	RN50	X
389	0.5x_resnet-18	X
390	resnet18-contrastive_multiview	X
391	resnet-18-LC_conv_init_m	X
392	resnet-18-LC_1st_conv_m	X
393	resnet18-supervised	X
394	resnet-18-LC_m	X
395	yudixie_resnet18_translation_reg_0_240719	X
396	r3m_resnet50_nocrop	X
397	yudixie_resnet18_distance_reg_0_240719	X
398	yudixie_resnet18_rotation_reg_0_240719	X
399	ReAlnet03_cornet	X
400	r3m_resnet34	X
401	ReAlnet02_cornet	X
402	resnet18_ecoset_full	X
403	resnet34_ecoset_full	X
404	r3m_resnet50	X
405	VOneCORnet-S	X
406	resnet50_ecoset_full	X
407	resnet101_ecoset_full	X

Benchmark bibtex

@article{geirhos2021partial,
              title={Partial success in closing the gap between human and machine vision},
              author={Geirhos, Robert and Narayanappa, Kantharaju and Mitzkus, Benjamin and Thieringer, Tizian and Bethge, Matthias and Wichmann, Felix A and Brendel, Wieland},
              journal={Advances in Neural Information Processing Systems},
              volume={34},
              year={2021},
              url={https://openreview.net/forum?id=QkljT4mrfs}
        }

Ceiling

0.39.

Note that scores are relative to this ceiling.

Data: Geirhos2021eidolonI

Metric: error_consistency