Sample stimuli

How to use

from brainscore_vision import load_benchmark
benchmark = load_benchmark("Geirhos2021eidolonII-error_consistency")
score = benchmark(my_model)

Benchmark API

Code examples

Model scores

Score Legend Min Alignment Max Alignment

Rank	Model	Score
1	SWSL_resnext101_32x8d	.732
2	effnetb1_cutmixrespatch_SAM_robust32_e10_manylayers_324x288	.703
3	ViT_B_16_imagenet1k	.699
4	vision_transformer_vit_large_patch16_224	.696
5	effnetb1_cutmixrespatch_SAM_robust32_e2_manylayers_324x288	.692
6	vit_base_patch16_clip_224:openai_ft_in1k	.680
7	vit_relpos_base_patch16_clsgap_224:sw_in1k	.674
8	convnext_tiny_imagenet_full_seed-0	.665
9	resnext101_32x48d_wsl	.664
10	effnetb1_cutmixpatch_SAM_robust32_avge6e8e9e10_manylayers_324x288	.662
11	resnext101_32x16d_wsl	.662
12	cvt_cvt-13-384-in22k_finetuned-in1k_4_LucyV4	.659
13	AdvProp_efficientnet-b6	.656
14	resnext101_32x32d_wsl	.650
15	cvt_cvt-21-384-in22k_finetuned-in1k_4_LucyV4	.650
16	resnet_152_v2	.646
17	focalnet_tiny_lrf_in1k	.646
18	cvt_cvt-13-224-in1k_4_LucyV4	.643
19	pnasnet_large	.640
20	vit_relpos_base_patch32_plus_rpn_256:sw_in1k	.635
21	convnext_small_imagenet_100_seed-0	.632
22	resnet_50_v2	.630
23	resnet50_finetune_cutmix_e3_robust_linf8255_e0_247x234	.626
24	vit_base_patch16_clip_224:openai_ft_in12k_in1k	.625
25	resnext101_32x8d_wsl	.624
26	CLIP_ViT-B_32	.622
27	ViT_L_32_imagenet1k	.617
28	convnext_large:fb_in22k_ft_in1k	.613
29	vit_large_patch14_clip_224:laion2b_ft_in1k	.613
30	SWSL_resnet50	.613
31	resnet_101_v2	.610
32	vit_large_patch14_clip_224:openai_ft_in1k	.603
33	resnet-152_v2-tf	.602
34	effnetb1_cutmixpatch_augmix_robust32_e5_324x288	.597
35	effnetb1_cutmixpatch_augmix_robust32_avge4e7_manylayers_324x288	.590
36	convnext_femto_ols:d1_in1k	.588
37	cvt_cvt-13-384-in1k_4_LucyV4	.582
38	AdvProp_efficientnet-b7	.578
39	pnasnet_large-tf	.577
40	convnext_large_imagenet_full_seed-0	.569
41	vit_huge_patch14_clip_224:laion2b_ft_in12k_in1k	.564
42	AT_efficientnet-b4	.559
43	AdvProp_efficientnet-b4	.559
44	CLIP_resnet50	.558
45	CLIP_resnet50_float32	.558
46	convnext_base:clip_laiona_augreg_ft_in1k_384	.555
47	fixres_resnext101_32x48d_wsl	.549
48	convnext_tiny:in12k_ft_in1k	.547
49	cvt_cvt-w24-384-in22k_finetuned-in1k_4	.546
50	effnetv2m_custom384	.540
51	efficientnet_b0	.538
52	cvt_cvt-21-384-in1k_4_LucyV4	.535
53	efficientnet-b7	.534
54	nasnet_large	.534
55	resnet152_imagenet_full	.527
56	resnet_152_v1	.524
57	resnet50-VITO-8deg-cc	.522
58	nasnet_large-tf	.520
59	cvt_cvt-21-224-in1k_4_LucyV4	.519
60	ViT_B_32_imagenet1k	.518
61	antialias-resnet152	.515
62	convnext_large_mlp:clip_laion2b_augreg_ft_in1k_384	.513
63	vit_huge_patch14_clip_336:laion2b_ft_in12k_in1k	.512
64	vit_large_patch14_clip_224:laion2b_ft_in12k_in1k	.509
65	vit_large_patch14_clip_336:openai_ft_in12k_in1k	.503
66	AT_efficientnet-b2	.501
67	AdvProp_efficientnet-b2	.501
68	deit_large_imagenet_full_seed-0	.496
69	effnetb1_272x240	.495
70	convnext_xlarge:fb_in22k_ft_in1k	.491
71	inception_v4	.482
72	vit_large_patch14_clip_336:laion2b_ft_in1k	.481
73	CLIP-RN50	.477
74	convnext_base_imagenet_full_seed-0	.475
75	vit_large_patch14_clip_224:openai_ft_in12k_in1k	.473
76	resnet_101_v1	.470
77	antialiased-rnext101_32x8d	.469
78	convnext_small_imagenet_full_seed-0	.465
79	deit_base_imagenet_full_seed-0	.465
80	inception_v3_pytorch	.463
81	inception_v3	.458
82	resnet-152_v1-tf	.455
83	grcnn_109	.454
84	tv_efficientnet-b1	.454
85	resnet50-moclr8deg	.453
86	resnet-152_v2_pytorch	.453
87	resnet101	.451
88	resnet50_imagenet_full	.448
89	pnasnet_large_pytorch	.445
90	cv_18_dagger_408_pretrained	.443
91	resnet101_imagenet_full	.433
92	densenet-201-keras	.423
93	inception_v4-tf	.423
94	densenet-201	.421
95	densenet_201_pytorch	.421
96	swin_small_patch4_window7_224:ms_in22k_ft_in1k	.416
97	resnet-50_v2-tf	.416
98	grcnn	.414
99	deit_small_imagenet_full_seed-0	.410
100	resnet50_random_linf8_perturb	.405
101	custom_model_cv_18_dagger_408	.401
102	xception-keras	.399
103	BiT-S-R50x3	.399
104	Res2Net50_26w_4s	.394
105	effnetb1_cutmix_augmix_sam_e1_5avg_424x377	.379
106	resnet34_imagenet_full	.379
107	inception_v3-tf	.375
108	BiT-S-R152x2	.371
109	resnet50_random_l2_perturb	.371
110	ReAlnet03	.366
111	vit_tiny_r_s16_p8_384:augreg_in21k_ft_in1k	.365
112	resnet50_moco_v2	.364
113	resnet50	.363
114	cornet_s	.361
115	convnext_xxlarge:clip_laion2b_soup_ft_in1k	.361
116	vonegrcnn_62e_nobn	.359
117	efficientnet_b2_imagenet_full	.359
118	resnet50_linf_4_robust	.358
119	vonegrcnn_47e	.354
120	vonegrcnn_52e_full	.351
121	AdvProp_efficientnet-b8	.350
122	ReAlnet01	.345
123	ReAlnet04	.342
124	resnet50_byol	.337
125	BiT-S-R101x3	.331
126	vonegrcnn_47e	.330
127	resnet50_tutorial	.329
128	resnet_50_v1	.329
129	resnet50-sup	.329
130	resnet-50-pytorch	.329
131	resnet18_imagenet_full	.329
132	vonegrcnn_52e_full	.325
133	densenet-121-keras	.320
134	resnet-50_v1-tf	.317
135	vgg_19	.317
136	antialiased-r50	.315
137	alexnet_random_linf8_perturb	.314
138	resnet_SIN_IN_FT_IN	.313
139	resnet50-SIN_IN_IN	.313
140	ReAlnet06	.313
141	densenet-169-keras	.311
142	vonegrcnn_62e_nobn	.311
143	alexnet_reduced_aliasing_early_checkpoint	.308
144	CORnet-S	.304
145	densenet-121	.302
146	ReAlnet05	.299
147	ReAlnet07	.298
148	xception	.298
149	resnet_50_v1_spiking	.297
150	voneresnet-50-robust	.296
151	mobilenet_v2_1.4_224-tf	.296
152	konkle_alexnetgn_ipcl_ref12_supervised_ipcl_aug	.294
153	inception_v2-tf	.294
154	voneresnet-50	.289
155	inception_v1-tf	.289
156	mobilenet_v2_1_4_224	.287
157	mobilenet_v2_1-4_224_pytorch	.287
158	voneresnet-50-non_stochastic	.286
159	resnet-101_v1-tf	.285
160	ReAlnet10	.279
161	ReAlnet08	.278
162	ReAlnet02	.274
163	densenet-169	.270
164	vgg_16	.270
165	mobilenet_v2_1_0_224	.267
166	resnet50_l2_3_robust	.266
167	mobilenet_v2_1_0_160	.266
168	ReAlnet09	.265
169	mobilenet_v2_1_3_224	.260
170	alexnet	.260
171	resnet50-SIN_IN	.258
172	BiT-S-R152x4	.257
173	BiT-S-R50x1	.256
174	alexnet_l2_3_robust	.255
175	resnet50-vitoimagevidnet8	.254
176	alexnet	.252
177	alexnet	.252
178	alexnet	.252
179	alexnet-baseline	.252
180	alexnet	.252
181	alexnet	.252
182	alexnet	.252
183	alexnet	.252
184	alexnet	.252
185	alexnet	.252
186	alexnet_ks_torevert	.252
187	nasnet_mobile	.250
188	blt_vs	.247
189	alexnet_linf_8_robust	.246
190	alexnet_robust_correct	.236
191	vonealexnet_gaussian_noise_std4_fixed	.235
192	inception_v1	.235
193	mobilenet_v2_0_5_224	.226
194	resnet50_robust_l2_eps3	.226
195	imagenet_l2_3_0	.226
196	resnet-50-robust	.226
197	mobilenet_v2_1.0_160-tf	.225
198	deit_small_imagenet_100_seed-0	.224
199	efficientnet_b0_imagenet_full	.224
200	konkle_alexnetgn_ipcl_ref01_primary_model	.223
201	convnext_tiny_sup	.223
202	resnet-18	.216
203	resnet-34	.216
204	alexnet_random_l2_3_perturb	.215
205	resnet50_robust_l2_eps1	.206
206	mobilenet_v2_1_0_192	.206
207	vgg-16-keras	.204
208	mobilenet_v2_1.3_224-tf	.198
209	mobilenet_v1_1.0_224	.196
210	resnet50_linf_8_robust	.195
211	BiT-S-R101x1	.191
212	regnet_y_400mf	.190
213	efficientnet_b1_imagenet_full	.189
214	mobilenet_v2_1.0_192-tf	.189
215	deit_small_imagenet_10_seed-0	.182
216	resnet50_imagenet_10_seed-0	.182
217	mobilenet_v1_0.75_224	.180
218	mobilenet_v2_1.0_224-tf	.180
219	resnet50-barlow	.179
220	resnet50-vicreg	.178
221	texture_shape_resnet50_trained_on_SIN	.178
222	deit_small_imagenet_1_seed-0	.174
223	nasnet_mobile-tf	.169
224	mobilenet_v2_0_5_192	.166
225	convnext_small_imagenet_10_seed-0	.161
226	mobilenet_v2_0.5_192-tf	.161
227	resnet50_simclr	.158
228	shufflenet_v2_x1_0	.158
229	mobilenet_v2_0_75_160	.155
230	mobilenet_v1_1.0_192	.154
231	alexnet_early_checkpoint	.151
232	imagenet_l2_10_0	.149
233	mobilenet_v1_0.75_128	.145
234	mobilenet_v2_0.5_224-tf	.143
235	alexnet_training_seed_01	.143
236	alexnet_training_seed_05	.143
237	0.5x_resnet-18_LC_w_sh_1_iter	.142
238	mobilenet_v2_0.35_160	.137
239	vgg-19-keras	.136
240	resnet50-SIN	.134
241	mobilenet_v2_0.5_96	.133
242	tf_efficientnetv2_s_in21ft1k_robust_linf12255_400x400	.128
243	texture_shape_alexnet_trained_on_SIN	.124
244	mobilenet_v2_0.75_192-tf	.124
245	AdvProp_efficientnet-b0	.118
246	resnet50_imagenet_100_seed-0	.118
247	mobilenet_v1_0.75_192	.116
248	AlexNet_SIN	.115
249	mobilenet_v2_0.75_224-tf	.114
250	mobilenet_v2_0.5_160	.114
251	mobilenet_v1_0.25_160	.112
252	mobilenet_v2_0_75_192	.110
253	mobilenet_v2_0.35_128	.109
254	mobilenet_v2_0_75_224	.107
255	mobilenet_v2_0.75_128	.105
256	mobilenet_v2_1_0_128	.105
257	mobilenet_v2_0.35_96	.104
258	mobilenet_v1_0.75_160	.100
259	alexnet_training_seed_06	.098
260	mobilenet_v1_1.0_160	.098
261	mobilenet_v1_0.5_192	.096
262	mobilenet_v2_1.0_96	.089
263	mobilenet_v2_1.0_128-tf	.088
264	mobilenet_v2_0.75_160-tf	.086
265	mobilenet_v1_0.5_224	.086
266	mobilenet_v1_0.25_192	.086
267	mobilenet_v2_0.35_192	.078
268	squeezenet1_1	.075
269	mobilenet_v1_1.0_128	.071
270	my-model	.065
271	mobilenet_v2_0.75_96	.065
272	resnet_50_v1_spiking_l4a	.061
273	mobilenet_v2_0.35_224	.061
274	squeezenet1_0	.058
275	resnet50-simclr-vissl	.058
276	resnet50-vicregl0p9	.057
277	mobilevit_small	.052
278	alexnet_training_seed_10	.049
279	alexnet_training_seed_04	.049
280	alexnet_training_seed_08	.049
281	alexnet_training_seed_02	.049
282	resnet50-vicregl0p75	.048
283	mobilenet_v1_0.25_128	.046
284	resnet50-meshes-lt-100-original-scratch	.043
285	mobilenet_v2_0.5_128	.042
286	resnet-50_untrained	.041
287	mobilenet_v1_0.5_160	.040
288	mobilenet_v1_0.25_224	.039
289	convnext_small_imagenet_1_seed-0	.032
290	alexnet_training_seed_03	.030
291	resnet50-meshes-lt-100-original-pretrained	.030
292	resnet_50_v1_spiking_l3	.026
293	resnet50_FractalDB	.024
294	alexnet_training_seed_07	.024
295	resnet_50_v1_spiking_l4	.020
296	alexnet_training_seed_09	.013
297	resnet50_imagenet_1_seed-0	.010
298	mobilenet_v1_0.5_128	.007
299	artResNet18_1	.006
300	barlow-twins-resnet50	.002
301	resnet-18-LC_1st_conv_m
302	resnet18_ecoset_full
303	resnet34_ecoset_full
304	resnet18-supervised
305	VOneCORnet-S
306	resnet50_ecoset_full
307	resnet101_ecoset_full
308	yudixie_resnet50_distance_rotation_0_240908
309	ViT-B/32
310	yudixie_resnet50_rotation_reg_0_240908
311	ReAlnet04_cornet
312	resnet152_ecoset_full
313	r3m_resnet50_nocrop
314	resnet-18-LC_m
315	yudixie_resnet50_distance_translation_0_240908
316	yudixie_resnet18_distance_translation_0_240719
317	r3m_resnet34
318	yudixie_resnet18_translation_rotation_0_240719
319	yudixie_resnet18_distance_rotation_0_240719
320	yudixie_resnet18_category_class_0_240719
321	ReAlnet05_cornet
322	yudixie_resnet18_random_0_240719
323	r3m_resnet50
324	resnet-18_test_m
325	yudixie_resnet18_cat_obj_class_all_latents_0_240719
326	yudixie_resnet18_distance_translation_rotation_0_240719
327	yudixie_resnet18_object_class_0_240719
328	ReAlnet06_cornet
329	yudixie_resnet50_translation_reg_0_240908
330	yudixie_resnet50_translation_rotation_0_240908
331	ReAlnet07_cornet
332	ReAlnet08_cornet
333	resnet18-colorization
334	0.5x_resnet-18_LC_w_sh_10_iter
335	omnivore_swinB
336	vggface
337	my-model
338	ReAlnet09_cornet
339	resnet18-instance_recognition
340	CORnet-Z
341	omnivore_swinS
342	ReAlnet10_cornet
343	r3m_resnet18
344	resnet18-deepcluster
345	yudixie_resnet50_distance_translation_rotation_0_240908
346	CORnetZ_CIFAR10
347	0.5x_resnet-18_LC
348	resnet18-local_aggregation
349	resnet-18-LC_w_sh_100_iter_conv_init_m
350	FrankRobWobv0
351	resnet50_primary_visual_cortex
352	resnet-18-LC_w_sh_10_iter_conv_init_m
353	resnet-18-LC_w_sh_100_iter_m
354	BiT-M-R152x2
355	resnet-18-LC_w_sh_1_iter_m
356	resnet-18-LC_w_sh_1_iter_conv_init_m
357	fulltest_microblockvf_nobottleneck_freeat_nopretrain_eps2_m4
358	resnet50_ImageNet
359	yudixie_resnet50_random_0_240908
360	resnet50_pretrained_with_retinal_waves
361	yudixie_resnet50_cat_obj_class_all_latents_0_240908
362	resnet50-cifar
363	yudixie_resnet50_imagenet1kpret_0_240312
364	yudixie_resnet50_object_class_0_240908
365	openclip
366	resnet50
367	dorinet_cornet_z
368	pixels
369	CORnetZ_CIFAR10_bs32_20_04
370	yudixie_resnet50_imagenet1kpret_0_240908
371	unet_entire
372	v1-pyr-nodown
373	dcgan
374	resnet-50x2_untrained
375	yudixie_resnet50_category_class_0_240908
376	resnet-18_untrained
377	resnet-18-LC_untrained
378	omnivore_swinT
379	resnet18_imagenet21kP
380	yudixie_resnet50_distance_reg_0_240908
381	resnet18-contrastive_predictive
382	ConvLSTM
383	r3m_resnet34_nocrop
384	PredRNN
385	SimVP
386	TAU
387	MIM
388	hmax
389	ViT_B_16
390	resnet18-depth_prediction
391	ViT_B_32
392	ViT_L_32
393	ReAlnet01_cornet
394	briaai_rmbg_1_4
395	resnet18-simclr
396	resnet18-autoencoder
397	resnet18-relative_position
398	BiT-M-R50x1
399	RN50
400	BiT-M-R50x3
401	0.5x_resnet-18_LC_w_sh_100_iter
402	RN50
403	yolos_tiny
404	resnet-18-LC_w_sh_10_iter_m
405	0.5x_resnet-18
406	yudixie_resnet18_translation_reg_0_240719
407	yudixie_resnet18_distance_reg_0_240719
408	ReAlnet03_cornet
409	yudixie_resnet18_rotation_reg_0_240719
410	yudixie_resnet18_imagenet1kpret_0_240719
411	ReAlnet02_cornet
412	resnet18-contrastive_multiview
413	resnet-18-LC_conv_init_m

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

0.45.

Note that scores are relative to this ceiling.

Data: Geirhos2021eidolonII

Metric: error_consistency