Sample stimuli

How to use

from brainscore_vision import load_benchmark
benchmark = load_benchmark("Geirhos2021phasescrambling-error_consistency")
score = benchmark(my_model)

Benchmark API

Code examples

Model scores

Score Legend Min Alignment Max Alignment

Rank	Model	Score
1	cvt_cvt-21-384-in1k_4_LucyV4	.734
2	cvt_cvt-w24-384-in22k_finetuned-in1k_4	.730
3	vit_large_patch14_clip_224:openai_ft_in12k_in1k	.677
4	convnext_xxlarge:clip_laion2b_soup_ft_in1k	.624
5	convnext_base:clip_laiona_augreg_ft_in1k_384	.621
6	vit_large_patch14_clip_224:laion2b_ft_in1k	.612
7	convnext_large:fb_in22k_ft_in1k	.606
8	vit_huge_patch14_clip_224:laion2b_ft_in12k_in1k	.600
9	vit_large_patch14_clip_336:laion2b_ft_in1k	.599
10	cvt_cvt-13-384-in1k_4_LucyV4	.596
11	cvt_cvt-21-384-in22k_finetuned-in1k_4_LucyV4	.584
12	vit_large_patch14_clip_224:openai_ft_in1k	.576
13	vit_large_patch14_clip_224:laion2b_ft_in12k_in1k	.576
14	resnet_152_v2	.572
15	fixres_resnext101_32x48d_wsl	.567
16	vit_base_patch16_clip_224:openai_ft_in12k_in1k	.562
17	convnext_large_mlp:clip_laion2b_augreg_ft_in1k_384	.561
18	vit_huge_patch14_clip_336:laion2b_ft_in12k_in1k	.521
19	resnet-152_v2_pytorch	.512
20	SWSL_resnext101_32x8d	.511
21	vit_large_patch14_clip_336:openai_ft_in12k_in1k	.511
22	vision_transformer_vit_large_patch16_224	.506
23	vit_base_patch16_clip_224:openai_ft_in1k	.503
24	convnext_xlarge:fb_in22k_ft_in1k	.499
25	resnext101_32x32d_wsl	.478
26	resnext101_32x48d_wsl	.462
27	convnext_tiny:in12k_ft_in1k	.459
28	resnet_50_v2	.448
29	vit_relpos_base_patch16_clsgap_224:sw_in1k	.442
30	resnext101_32x16d_wsl	.431
31	SWSL_resnet50	.419
32	effnetv2m_custom384	.416
33	CLIP_resnet50_float32	.404
34	CLIP_resnet50	.404
35	effnetb1_cutmixpatch_SAM_robust32_avge6e8e9e10_manylayers_324x288	.383
36	cvt_cvt-13-384-in22k_finetuned-in1k_4_LucyV4	.382
37	effnetb1_cutmixrespatch_SAM_robust32_e10_manylayers_324x288	.378
38	pnasnet_large	.376
39	convnext_tiny_imagenet_full_seed-0	.372
40	CLIP_ViT-B_32	.367
41	resnet_101_v2	.364
42	ViT_B_16_imagenet1k	.363
43	resnext101_32x8d_wsl	.363
44	pnasnet_large-tf	.361
45	convnext_femto_ols:d1_in1k	.350
46	convnext_base_imagenet_full_seed-0	.347
47	focalnet_tiny_lrf_in1k	.344
48	cvt_cvt-13-224-in1k_4_LucyV4	.342
49	convnext_large_imagenet_full_seed-0	.339
50	effnetb1_cutmixrespatch_SAM_robust32_e2_manylayers_324x288	.338
51	CLIP-RN50	.331
52	resnet50_imagenet_full	.320
53	inception_v4	.312
54	nasnet_large	.308
55	effnetb1_272x240	.307
56	resnet152_imagenet_full	.289
57	convnext_small_imagenet_full_seed-0	.289
58	cvt_cvt-21-224-in1k_4_LucyV4	.289
59	cv_18_dagger_408_pretrained	.286
60	resnet101	.273
61	resnet50-VITO-8deg-cc	.269
62	ViT_L_32_imagenet1k	.268
63	inception_v3	.267
64	swin_small_patch4_window7_224:ms_in22k_ft_in1k	.256
65	inception_v3-tf	.251
66	resnet101_imagenet_full	.250
67	AT_efficientnet-b4	.250
68	AdvProp_efficientnet-b4	.250
69	effnetb1_cutmixpatch_augmix_robust32_avge4e7_manylayers_324x288	.249
70	ViT_B_32_imagenet1k	.247
71	resnet-152_v2-tf	.246
72	effnetb1_cutmixpatch_augmix_robust32_e5_324x288	.223
73	vit_tiny_r_s16_p8_384:augreg_in21k_ft_in1k	.222
74	xception-keras	.222
75	resnet50_finetune_cutmix_e3_robust_linf8255_e0_247x234	.219
76	inception_v4-tf	.214
77	resnet_152_v1	.210
78	nasnet_large-tf	.209
79	deit_large_imagenet_full_seed-0	.207
80	antialias-resnet152	.202
81	inception_v3_pytorch	.195
82	resnet34_imagenet_full	.192
83	convnext_small_imagenet_100_seed-0	.189
84	resnet_101_v1	.183
85	resnet50-moclr8deg	.182
86	vit_relpos_base_patch32_plus_rpn_256:sw_in1k	.181
87	effnetb1_cutmix_augmix_sam_e1_5avg_424x377	.179
88	Res2Net50_26w_4s	.177
89	resnet50	.170
90	AdvProp_efficientnet-b7	.168
91	densenet-201-keras	.165
92	pnasnet_large_pytorch	.162
93	AdvProp_efficientnet-b2	.161
94	AT_efficientnet-b2	.161
95	resnet50_random_l2_perturb	.159
96	resnet-50_v2-tf	.158
97	deit_base_imagenet_full_seed-0	.156
98	efficientnet_b0	.154
99	densenet_201_pytorch	.153
100	densenet-201	.153
101	resnet50_random_linf8_perturb	.151
102	xception	.150
103	grcnn_109	.148
104	vonegrcnn_47e	.147
105	deit_small_imagenet_100_seed-0	.146
106	ReAlnet_incorrect	.145
107	ReAlnet01	.145
108	BiT-S-R152x2	.144
109	resnet18_imagenet_full	.142
110	custom_model_cv_18_dagger_408	.138
111	antialiased-rnext101_32x8d	.137
112	antialiased-r50	.134
113	densenet-169	.133
114	AdvProp_efficientnet-b6	.130
115	ReAlnet07	.129
116	AdvProp_efficientnet-b8	.125
117	inception_v2-tf	.123
118	alexnet_random_l2_3_perturb	.123
119	densenet-169-keras	.120
120	grcnn	.120
121	alexnet_random_linf8_perturb	.119
122	resnet50-sup	.118
123	resnet-50-pytorch	.118
124	resnet50_tutorial	.118
125	resnet_50_v1	.118
126	ReAlnet03	.116
127	vonegrcnn_62e_nobn	.114
128	resnet50-meshes-lt-100-original-pretrained	.114
129	resnet-50_untrained	.114
130	deit_small_imagenet_1_seed-0	.114
131	resnet50_imagenet_10_seed-0	.114
132	deit_small_imagenet_10_seed-0	.114
133	alexnet_linf_8_robust	.112
134	ReAlnet05	.112
135	resnet50-vicregl0p9	.111
136	efficientnet-b7	.111
137	vonegrcnn_52e_full	.110
138	vonegrcnn_47e	.110
139	cornet_s	.109
140	ReAlnet06	.108
141	resnet50_moco_v2	.108
142	alexnet_training_seed_01	.107
143	alexnet_training_seed_05	.107
144	vonegrcnn_62e_nobn	.104
145	BiT-S-R50x3	.101
146	convnext_tiny_sup	.101
147	artResNet18_1	.100
148	resnet50_linf_4_robust	.100
149	nasnet_mobile-tf	.097
150	resnet-34	.094
151	resnet-101_v1-tf	.094
152	ReAlnet04	.094
153	ReAlnet10	.094
154	tf_efficientnetv2_s_in21ft1k_robust_linf12255_400x400	.093
155	konkle_alexnetgn_ipcl_ref12_supervised_ipcl_aug	.091
156	resnet50-barlow	.091
157	resnet50-SIN_IN	.091
158	resnet50-vitoimagevidnet8	.090
159	ReAlnet08	.090
160	deit_small_imagenet_full_seed-0	.089
161	imagenet_l2_10_0	.089
162	alexnet_l2_3_robust	.087
163	alexnet	.084
164	vonegrcnn_52e_full	.084
165	convnext_small_imagenet_10_seed-0	.083
166	vgg_19	.083
167	alexnet_early_checkpoint	.081
168	densenet-121-keras	.081
169	voneresnet-50-robust	.080
170	alexnet_robust_correct	.080
171	ReAlnet02	.080
172	voneresnet-50	.080
173	resnet_SIN_IN_FT_IN	.079
174	resnet50-SIN_IN_IN	.079
175	mobilenet_v2_1_3_224	.079
176	resnet-18	.077
177	resnet50_imagenet_100_seed-0	.076
178	densenet-121	.075
179	resnet50_robust_l2_eps1	.074
180	resnet50_linf_8_robust	.073
181	vgg-19-keras	.069
182	inception_v1-tf	.069
183	ReAlnet09	.069
184	resnet50-simclr-vissl	.068
185	mobilenet_v2_0.5_96	.067
186	texture_shape_resnet50_trained_on_SIN	.067
187	convnext_small_imagenet_1_seed-0	.066
188	barlow-twins-resnet50	.066
189	resnet-50_v1-tf	.066
190	mobilenet_v2_0.35_96	.063
191	resnet50-vicregl0p75	.060
192	alexnet_training_seed_02	.060
193	alexnet_training_seed_10	.060
194	alexnet_training_seed_04	.060
195	alexnet_training_seed_08	.060
196	alexnet_ks_torevert	.060
197	alexnet	.060
198	alexnet	.060
199	alexnet	.060
200	alexnet	.060
201	alexnet	.060
202	alexnet	.060
203	alexnet	.060
204	alexnet-baseline	.060
205	alexnet	.060
206	alexnet	.060
207	resnet50_l2_3_robust	.060
208	voneresnet-50-non_stochastic	.059
209	resnet50_simclr	.059
210	mobilenet_v1_0.25_128	.057
211	BiT-S-R101x1	.057
212	inception_v1	.055
213	BiT-S-R152x4	.055
214	efficientnet_b2_imagenet_full	.054
215	vgg-16-keras	.052
216	mobilenet_v2_0_5_224	.052
217	CORnet-S	.051
218	mobilevit_small	.051
219	nasnet_mobile	.051
220	resnet50_byol	.050
221	mobilenet_v2_1_0_224	.049
222	mobilenet_v2_1-4_224_pytorch	.048
223	mobilenet_v2_1_4_224	.048
224	resnet50-SIN	.048
225	mobilenet_v2_0.35_192	.047
226	mobilenet_v2_1.0_224-tf	.047
227	konkle_alexnetgn_ipcl_ref01_primary_model	.045
228	texture_shape_alexnet_trained_on_SIN	.045
229	regnet_y_400mf	.044
230	vgg_16	.044
231	alexnet_training_seed_03	.042
232	alexnet_training_seed_07	.041
233	mobilenet_v2_1.0_192-tf	.040
234	mobilenet_v1_1.0_160	.040
235	alexnet_training_seed_06	.037
236	squeezenet1_1	.034
237	resnet50-meshes-lt-100-original-scratch	.033
238	mobilenet_v2_0_75_192	.033
239	mobilenet_v1_0.75_128	.033
240	mobilenet_v2_0.5_128	.032
241	mobilenet_v2_1.0_96	.032
242	AdvProp_efficientnet-b0	.031
243	alexnet_reduced_aliasing_early_checkpoint	.031
244	mobilenet_v2_0.75_160-tf	.030
245	blt_vs	.030
246	mobilenet_v2_0.75_96	.029
247	mobilenet_v2_1_0_192	.029
248	imagenet_l2_3_0	.029
249	resnet-50-robust	.029
250	resnet50_robust_l2_eps3	.029
251	alexnet_training_seed_09	.028
252	mobilenet_v2_1.3_224-tf	.028
253	mobilenet_v2_0.35_224	.027
254	AlexNet_SIN	.027
255	mobilenet_v1_0.75_160	.025
256	mobilenet_v1_0.25_224	.024
257	resnet-152_v1-tf	.023
258	mobilenet_v1_1.0_128	.023
259	resnet50_imagenet_1_seed-0	.022
260	vonealexnet_gaussian_noise_std4_fixed	.022
261	mobilenet_v2_0_75_224	.020
262	mobilenet_v1_0.75_192	.019
263	mobilenet_v2_1_0_128	.019
264	efficientnet_b0_imagenet_full	.019
265	mobilenet_v2_0_75_160	.018
266	resnet50-vicreg	.017
267	mobilenet_v2_1.0_160-tf	.017
268	BiT-S-R50x1	.016
269	shufflenet_v2_x1_0	.016
270	mobilenet_v1_0.5_128	.015
271	0.5x_resnet-18_LC_w_sh_1_iter	.014
272	mobilenet_v2_0_5_192	.013
273	squeezenet1_0	.013
274	mobilenet_v2_0.75_128	.012
275	resnet50_FractalDB	.012
276	mobilenet_v1_0.25_160	.011
277	mobilenet_v1_1.0_224	.010
278	mobilenet_v2_0.5_192-tf	.009
279	mobilenet_v1_0.5_192	.008
280	BiT-S-R101x3	.007
281	mobilenet_v2_1_0_160	.007
282	mobilenet_v2_1.4_224-tf	.006
283	efficientnet_b1_imagenet_full	.006
284	mobilenet_v2_0.75_224-tf	.006
285	mobilenet_v1_0.5_160	.005
286	mobilenet_v1_1.0_192	.005
287	mobilenet_v2_0.35_128	.005
288	mobilenet_v1_0.75_224	.003
289	mobilenet_v2_0.5_224-tf	.003
290	mobilenet_v2_0.35_160	.002
291	mobilenet_v2_0.75_192-tf	.001
292	mobilenet_v2_0.5_160	.001
293	mobilenet_v1_0.5_224	.001
294	mobilenet_v1_0.25_192	.001
295	mobilenet_v2_1.0_128-tf	.000
296	resnet-18-LC_conv_init_m	X
297	resnet-18-LC_1st_conv_m	X
298	resnet18-supervised	X
299	r3m_resnet50_nocrop	X
300	resnet-18-LC_m	X
301	resnet18-colorization	X
302	r3m_resnet34	X
303	resnet-18-LC_w_sh_1_iter_m	X
304	CORnetZ_CIFAR10_bs32_20_04	X
305	yudixie_resnet50_imagenet1kpret_0_240312	X
306	yudixie_resnet18_distance_reg_0_240719	X
307	yudixie_resnet18_rotation_reg_0_240719	X
308	yudixie_resnet18_imagenet1kpret_0_240719	X
309	ReAlnet03_cornet	X
310	r3m_resnet50	X
311	ReAlnet02_cornet	X
312	BiT-M-R50x1	X
313	BiT-M-R50x3	X
314	resnet18_ecoset_full	X
315	resnet34_ecoset_full	X
316	resnet50-cifar	X
317	resnet50_ecoset_full	X
318	VOneCORnet-S	X
319	pixels	X
320	resnet101_ecoset_full	X
321	yudixie_resnet50_distance_rotation_0_240908	X
322	0.5x_resnet-18_LC_w_sh_10_iter	X
323	yudixie_resnet50_rotation_reg_0_240908	X
324	resnet152_ecoset_full	X
325	my-model	X
326	yudixie_resnet18_distance_translation_0_240719	X
327	ViT_B_16	X
328	yudixie_resnet50_distance_translation_0_240908	X
329	yudixie_resnet18_translation_rotation_0_240719	X
330	yudixie_resnet18_distance_rotation_0_240719	X
331	yudixie_resnet18_category_class_0_240719	X
332	ReAlnet05_cornet	X
333	yudixie_resnet18_random_0_240719	X
334	resnet-18_test_m	X
335	yudixie_resnet18_cat_obj_class_all_latents_0_240719	X
336	yudixie_resnet50_random_0_240908	X
337	vggface	X
338	yudixie_resnet18_distance_translation_rotation_0_240719	X
339	yudixie_resnet18_object_class_0_240719	X
340	yudixie_resnet18_translation_reg_0_240719	X
341	ReAlnet06_cornet	X
342	yudixie_resnet50_translation_reg_0_240908	X
343	ViT-B/32	X
344	yudixie_resnet50_translation_rotation_0_240908	X
345	ReAlnet07_cornet	X
346	dorinet_cornet_z	X
347	my-model	X
348	resnet-18-LC_w_sh_100_iter_conv_init_m	X
349	ReAlnet08_cornet	X
350	resnet50	X
351	openclip	X
352	omnivore_swinB	X
353	resnet18-instance_recognition	X
354	ReAlnet09_cornet	X
355	resnet-18-LC_w_sh_10_iter_conv_init_m	X
356	fulltest_microblockvf_nobottleneck_freeat_nopretrain_eps2_m4	X
357	CORnet-Z	X
358	omnivore_swinS	X
359	r3m_resnet18	X
360	ReAlnet10_cornet	X
361	resnet-18-LC_w_sh_100_iter_m	X
362	resnet18-deepcluster	X
363	resnet18-local_aggregation	X
364	yudixie_resnet50_distance_translation_rotation_0_240908	X
365	FrankRobWobv0	X
366	0.5x_resnet-18_LC	X
367	resnet50_ImageNet	X
368	CORnetZ_CIFAR10	X
369	resnet50_pretrained_with_retinal_waves	X
370	ReAlnet04_cornet	X
371	resnet18-depth_prediction	X
372	yudixie_resnet50_object_class_0_240908	X
373	r3m_resnet34_nocrop	X
374	ViT_B_32	X
375	resnet18-contrastive_predictive	X
376	omnivore_swinT	X
377	ViT_L_32	X
378	resnet18-simclr	X
379	resnet18-autoencoder	X
380	resnet-18-LC_untrained	X
381	resnet18-relative_position	X
382	resnet-18-LC_w_sh_1_iter_conv_init_m	X
383	yudixie_resnet50_category_class_0_240908	X
384	RN50	X
385	0.5x_resnet-18_LC_w_sh_100_iter	X
386	yudixie_resnet50_cat_obj_class_all_latents_0_240908	X
387	RN50	X
388	resnet-18_untrained	X
389	0.5x_resnet-18	X
390	resnet18_imagenet21kP	X
391	yudixie_resnet50_distance_reg_0_240908	X
392	ConvLSTM	X
393	resnet-18-LC_w_sh_10_iter_m	X
394	SimVP	X
395	TAU	X
396	MIM	X
397	hmax	X
398	yudixie_resnet50_imagenet1kpret_0_240908	X
399	resnet-50x2_untrained	X
400	dcgan	X
401	BiT-M-R152x2	X
402	ReAlnet01_cornet	X
403	resnet18-contrastive_multiview	X
404	unet_entire	X
405	briaai_rmbg_1_4	X
406	yolos_tiny	X
407	v1-pyr-nodown	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

0.45.

Note that scores are relative to this ceiling.

Data: Geirhos2021phasescrambling

Metric: error_consistency