Leaderboard Explore Compare Profile Community Blog FAQ Competition 2024 Report Issue

Sample stimuli

How to use

from brainscore_vision import load_benchmark
benchmark = load_benchmark("Kar2019-ost")
score = benchmark(my_model)

Benchmark API

Code examples

Model scores

Score Legend Min Alignment Max Alignment

Rank	Model	Score
1	ReAlnet04_cornet	.310
2	ReAlnet05_cornet	.306
3	CORnet-S	.305
4	VOneCORnet-S	.297
5	ReAlnet08_cornet	.290
6	ReAlnet03_cornet	.276
7	ReAlnet05_ALL	.273
8	ReAlnet07_cornet	.272
9	ReAlnet09_cornet	.270
10	ReAlnet01_cornet	.267
11	ReAlnet04_ALL	.267
12	ReAlnet08_ALL	.260
13	ReAlnet07_ALL	.259
14	ReAlnet02_ALL	.258
15	ReAlnet09_ALL	.255
16	ReAlnet10_cornet	.255
17	ReAlnet06_cornet	.252
18	ReAlnet03_ALL	.251
19	ReAlnet06_ALL	.247
20	ReAlnet10_ALL	.240
21	ReAlnet02_cornet	.233
22	ReAlnet01_ALL	.232
23	bagnet9	nan
24	nasnet_mobile-tf	nan
25	nasnet_large-tf	nan
26	pnasnet_large-tf	nan
27	alexnet	nan
28	CORnet-Z	nan
29	squeezenet1_0	nan
30	squeezenet1_1	nan
31	xception-keras	nan
32	densenet-121-keras	nan
33	densenet-169-keras	nan
34	densenet-201-keras	nan
35	inception_v1-tf	nan
36	inception_v2-tf	nan
37	inception_v3-tf	nan
38	inception_v4-tf	nan
39	inception_resnet_v2	nan
40	resnet-18	nan
41	resnet-34	nan
42	resnet-50-pytorch	nan
43	resnet-50_v1-tf	nan
44	resnet-101_v1-tf	nan
45	resnet-152_v1-tf	nan
46	resnet-50_v2-tf	nan
47	resnet-101_v2-tf	nan
48	resnet-152_v2-tf	nan
49	resnet-50-robust	nan
50	vgg-16-keras	nan
51	vgg-19-keras	nan
52	vggface	nan
53	mobilenet_v1_1.0_224	nan
54	mobilenet_v1_1.0_192	nan
55	mobilenet_v1_1.0_160	nan
56	mobilenet_v1_1.0_128	nan
57	mobilenet_v1_0.75_224	nan
58	mobilenet_v1_0.75_192	nan
59	mobilenet_v1_0.75_160	nan
60	mobilenet_v1_0.75_128	nan
61	mobilenet_v1_0.5_224	nan
62	mobilenet_v1_0.5_192	nan
63	mobilenet_v1_0.5_160	nan
64	mobilenet_v1_0.5_128	nan
65	mobilenet_v1_0.25_224	nan
66	mobilenet_v1_0.25_192	nan
67	mobilenet_v1_0.25_160	nan
68	mobilenet_v1_0.25_128	nan
69	mobilenet_v2_1.4_224-tf	nan
70	mobilenet_v2_1.3_224-tf	nan
71	mobilenet_v2_1.0_224-tf	nan
72	mobilenet_v2_1.0_192-tf	nan
73	mobilenet_v2_1.0_160-tf	nan
74	mobilenet_v2_1.0_128-tf	nan
75	mobilenet_v2_1.0_96	nan
76	mobilenet_v2_0.75_224-tf	nan
77	mobilenet_v2_0.75_192-tf	nan
78	mobilenet_v2_0.75_160-tf	nan
79	mobilenet_v2_0.75_128	nan
80	mobilenet_v2_0.75_96	nan
81	mobilenet_v2_0.5_224-tf	nan
82	mobilenet_v2_0.5_192-tf	nan
83	mobilenet_v2_0.5_160	nan
84	mobilenet_v2_0.5_128	nan
85	mobilenet_v2_0.5_96	nan
86	mobilenet_v2_0.35_224	nan
87	mobilenet_v2_0.35_192	nan
88	mobilenet_v2_0.35_160	nan
89	mobilenet_v2_0.35_128	nan
90	mobilenet_v2_0.35_96	nan
91	resnet50-SIN	nan
92	resnet50-SIN_IN	nan
93	resnet50-SIN_IN_IN	nan
94	resnext101_32x48d_wsl	nan
95	resnext101_32x8d_wsl	nan
96	resnext101_32x16d_wsl	nan
97	resnext101_32x32d_wsl	nan
98	fixres_resnext101_32x48d_wsl	nan
99	resnet18-instance_recognition	nan
100	resnet18-supervised	nan
101	pixels
102	resnet18-local_aggregation	nan
103	resnet18-colorization	nan
104	resnet18-relative_position	nan
105	resnet18-depth_prediction	nan
106	resnet18-contrastive_predictive	nan
107	resnet18-autoencoder	nan
108	AT_efficientnet-b4	nan
109	AT_efficientnet-b2	nan
110	AT_efficientnet-b7	nan
111	efficientnet-b0
112	efficientnet-b6
113	efficientnet-b2
114	efficientnet-b4
115	AT_efficientnet-b0
116	prednet
117	resnet18-deepcluster	nan
118	resnet18-contrastive_multiview	nan
119	ViT_L_32
120	ViT_B_32_imagenet1k	nan
121	ViT_L_16_imagenet1k
122	ViT_B_32	nan
123	ViT_B_16_imagenet1k
124	BiT-M-R152x2	nan
125	BiT-M-R152x4	nan
126	deit_tiny_patch16_224	nan
127	deit_small_patch16_224	nan
128	deit_base_patch16_224	nan
129	deit_tiny_distilled_patch16_224	nan
130	BiT-M-R50x1	nan
131	ViT_B_16	nan
132	deit_small_distilled_patch16_224	nan
133	deit_base_distilled_patch16_224	nan
134	deit_base_patch16_384	nan
135	deit_base_distilled_patch16_384	nan
136	BiT-M-R50x3	nan
137	cornetz_contrastive	nan
138	BiT-M-R101x1	nan
139	BiT-M-R101x3	nan
140	BiT-S-R50x1	nan
141	BiT-S-R50x3	nan
142	alexnet	nan
143	BiT-S-R101x1	nan
144	BiT-S-R101x3	nan
145	BiT-S-R152x2	nan
146	ViT-B/32	nan
147	RN50	nan
148	deit_tiny_patch16_224_id	nan
149	deit_small_patch16_224_id	nan
150	deit_base_patch16_224_id	nan
151	deit_tiny_distilled_patch16_224_id	nan
152	deit_small_distilled_patch16_224_id	nan
153	deit_base_distilled_patch16_224_id	nan
154	deit_base_patch16_384_id	nan
155	deit_base_distilled_patch16_384_id	nan
156	alexnet	nan
157	efficientnet-b7	nan
158	AdvProp_efficientnet-b0	nan
159	AdvProp_efficientnet-b2	nan
160	unet_entire	nan
161	resnet18-simclr	nan
162	AdvProp_efficientnet-b6	nan
163	AdvProp_efficientnet-b4	nan
164	AdvProp_efficientnet-b7	nan
165	AdvProp_efficientnet-b8	nan
166	barlow-twins-resnet50	nan
167	voneresnet-50-robust	nan
168	voneresnet-50-non_stochastic	nan
169	voneresnet-50	nan
170	effnetb1_cutmix_augmix_sam_e1_5avg_424x377	nan
171	CLIP-RN50	nan
172	0.5x_resnet-18	nan
173	0.5x_resnet-18_LC_w_sh_10_iter	nan
174	0.5x_resnet-18_LC	nan
175	0.5x_resnet-18_LC_w_sh_100_iter	nan
176	0.5x_resnet-18_LC_w_sh_1_iter	nan
177	FrankRobWobv0	nan
178	effnetb1_cutmixpatch_augmix_robust32_avge4e7_manylayers_324x288	nan
179	grcnn_robust_v1	nan
180	alexnet	nan
181	alexnet-baseline	nan
182	RN50	nan
183	effnetb1_cutmixpatch_augmix_robust32_e5_324x288	nan
184	effnetv2m_custom384	nan
185	grcnn_v2_text_noise	nan
186	alexnet	nan
187	efficientnet_b0	nan
188	grcnn	nan
189	effnetb1_272x240	nan
190	custom_model_cv_18_dagger_408	nan
191	tf_efficientnetv2_s_in21ft1k_robust_linf12255_400x400	nan
192	grcnn_109	nan
193	cv_18_dagger_408_pretrained	nan
194	effnetb1_cutmixpatch_SAM_robust32_avge6e8e9e10_manylayers_324x288	nan
195	resnet_SIN_IN_FT_IN	nan
196	resnet-18-LC_1st_conv_m	nan
197	resnet-18_test_m	nan
198	resnet-18-LC_w_sh_1_iter_m	nan
199	resnet-18-LC_m	nan
200	resnet-18-LC_w_sh_10_iter_m	nan
201	resnet-18-LC_conv_init_m	nan
202	resnet-18-LC_w_sh_1_iter_conv_init_m	nan
203	resnet-18-LC_w_sh_10_iter_conv_init_m	nan
204	resnet-18-LC_w_sh_100_iter_m	nan
205	resnet-18-LC_untrained	nan
206	v1-pyr-nodown	nan
207	resnet-18-LC_w_sh_100_iter_conv_init_m	nan
208	resnet-18-LC_d_w_sh_1x1_conv_init_m	nan
209	effnetb1_cutmixrespatch_SAM_robust32_e2_manylayers_324x288	nan
210	alexnet	nan
211	effnetb1_cutmixrespatch_SAM_robust32_e10_manylayers_324x288	nan
212	resnet-18_untrained	nan
213	resnet50_finetune_cutmix_e3_robust_linf8255_e0_247x234	nan
214	resnet50_finetune_cutmix_AVGe2e3_robust_linf8255_e0_247x234	nan
215	CORnetZ_CIFAR10	nan
216	my-model	nan
217	vonegrcnn_47e	nan
218	vonegrcnn_52e_full	nan
219	vonegrcnn_62e_nobn	nan
220	resnet-50_untrained	nan
221	resnet-50x2_untrained	nan
222	CORnetZ_CIFAR10_bs32_20_04	nan
223	r3m_resnet50	nan
224	r3m_resnet34	nan
225	r3m_resnet18	nan
226	r3m_resnet50_nocrop	nan
227	r3m_resnet34_nocrop	nan
228	fitvid_trained_on_physion	nan
229	resnet50-sup	nan
230	resnet50-barlow	nan
231	resnet50-simclr-vissl	nan
232	omnivore_swinT	nan
233	omnivore_swinS	nan
234	omnivore_swinB	nan
235	alexnet	nan
236	my-model	nan
237	resnet50	nan
238	alexnet_robust_correct	nan
239	openclip
240	resnet50-meshes-lt-100-original-pretrained	nan
241	resnet50-meshes-lt-100-original-scratch	nan
242	resnet50_byol	nan
243	cornet_s	nan
244	alexnet	nan
245	resnet50_simclr	nan
246	alexnet_early_checkpoint	nan
247	resnet50	nan
248	alexnet_l2_3_robust	nan
249	vgg_19	nan
250	resnet50_moco_v2	nan
251	alexnet_random_l2_3_perturb	nan
252	alexnet_linf_8_robust	nan
253	alexnet_random_linf8_perturb	nan
254	resnet50_l2_3_robust	nan
255	resnet50_linf_4_robust	nan
256	alexnet_reduced_aliasing_early_checkpoint	nan
257	resnet50_linf_8_robust	nan
258	resnet50_random_linf8_perturb	nan
259	vonealexnet_gaussian_noise_std4_fixed	nan
260	resnet50_random_l2_perturb	nan
261	texture_shape_resnet50_trained_on_SIN	nan
262	texture_shape_alexnet_trained_on_SIN	nan
263	resnet101	nan
264	CLIP_resnet50	nan
265	vision_transformer_vit_large_patch16_224	nan
266	CLIP_ViT-B_32	nan
267	SWSL_resnet50	nan
268	SWSL_resnext101_32x8d	nan
269	CLIP_resnet50_float32	nan
270	konkle_alexnetgn_ipcl_ref12_supervised_ipcl_aug	nan
271	resnet50-vicregl0p75	nan
272	konkle_alexnetgn_ipcl_ref01_primary_model	nan
273	resnet50-vicreg	nan
274	resnet50-vicregl0p9	nan
275	resnet50-cifar	nan
276	alexnet	nan
277	resnet50_pretrained_with_retinal_waves	nan
278	resnet50_FractalDB	nan
279	resnet50_ImageNet
280	resnet50-moclr8deg	nan
281	resnet50-vitoimagevidnet8	nan
282	imagenet_l2_3_0	nan
283	imagenet_l2_10_0	nan
284	fulltest_microblockvf_nobottleneck_freeat_nopretrain_eps2_m4	nan
285	convnext_tiny_sup	nan
286	antialiased-r50	nan
287	antialiased-rnext101_32x8d	nan
288	shufflenet_v2_x1_0	nan
289	antialias-resnet152	nan
290	cvt_cvt-w24-384-in22k_finetuned-in1k_4	nan
291	focalnet_tiny_lrf_in1k	nan
292	Res2Net50_26w_4s	nan
293	resnet50_robust_l2_eps1	nan
294	resnet50_robust_l2_eps3	nan
295	resnet50-VITO-8deg-cc	nan
296	alexnet	nan
297	regnet_y_400mf	nan
298	convnext_large:fb_in22k_ft_in1k	nan
299	convnext_small_imagenet_100_seed-0	nan
300	tv_efficientnet-b1	nan
301	alexnet_ks_torevert	nan
302	deit_small_imagenet_10_seed-0	nan
303	resnet_50_v2	nan
304	alexnet_training_seed_09	nan
305	resnet_152_v1
306	resnet34_imagenet_full	nan
307	yudixie_resnet50_imagenet1kpret_0_240312	nan
308	resnet_50_v1_spiking	nan
309	efficientnet_b0_imagenet_full	nan
310	convnext_large_mlp:clip_laion2b_augreg_ft_in1k_384	nan
311	mobilevit_small	nan
312	alexnet_training_seed_05	nan
313	mobilenet_v2_0_75_224	nan
314	cvt_cvt-21-384-in22k_finetuned-in1k_4_LucyV4	nan
315	mobilenet_v2_1_0_128	nan
316	resnet_50_v1_spiking_l4	nan
317	resnet50_ewconplus_wzc	nan
318	convnext_tiny:in12k_ft_in1k	nan
319	mobilenet_v2_1_0_160	nan
320	mobilenet_v2_1_0_192	nan
321	inception_v3	nan
322	alexnet_training_seed_06	nan
323	ReAlnet03	nan
324	resnet18_imagenet21kP	nan
325	yudixie_resnet50_category_class_0_240908	nan
326	vit_large_patch14_clip_224:laion2b_ft_in1k	nan
327	ReAlnet04	nan
328	mobilenet_v2_0_5_192	nan
329	ReAlnet05	nan
330	resnet_50_v1_spiking_l4a	nan
331	yudixie_resnet18_rotation_reg_0_240719	nan
332	convnext_small_imagenet_full_seed-0	nan
333	cvt_cvt-13-224-in1k_4_LucyV4	nan
334	resnet50_imagenet_100_seed-0	nan
335	alexnet_training_seed_02	nan
336	AlexNet_SIN	nan
337	densenet-201	nan
338	yudixie_resnet50_distance_reg_0_240908	nan
339	vit_large_patch14_clip_336:laion2b_ft_in1k	nan
340	alexnet_training_seed_04	nan
341	ConvLSTM
342	PredRNN
343	SimVP
344	TAU
345	MIM
346	vit_large_patch14_clip_336:openai_ft_in12k_in1k	nan
347	hmax
348	resnet101_imagenet_full	nan
349	ViT_L_32_imagenet1k	nan
350	mobilenet_v2_1-4_224_pytorch	nan
351	convnext_small_imagenet_10_seed-0	nan
352	resnet-152_v2_pytorch	nan
353	ReAlnet07	nan
354	vit_relpos_base_patch16_clsgap_224:sw_in1k	nan
355	densenet_201_pytorch
356	pnasnet_large_pytorch
357	ReAlnet06	nan
358	inception_v3_pytorch
359	efficientnet_b2_imagenet_full	nan
360	convnext_large_imagenet_full_seed-0	nan
361	ReAlnet08	nan
362	vit_relpos_base_patch32_plus_rpn_256:sw_in1k	nan
363	yolos_tiny	nan
364	briaai_rmbg_1_4	nan
365	deit_large_imagenet_full_seed-0	nan
366	cvt_cvt-13-384-in22k_finetuned-in1k_4_LucyV4	nan
367	efficientnet_b1_imagenet_full	nan
368	yudixie_resnet18_object_class_0_240719	nan
369	deit_base_imagenet_full_seed-0	nan
370	convnext_small_imagenet_1_seed-0	nan
371	deit_small_imagenet_100_seed-0	nan
372	convnext_base_imagenet_full_seed-0	nan
373	convnext_tiny_imagenet_full_seed-0	nan
374	deit_small_imagenet_full_seed-0	nan
375	vit_tiny_r_s16_p8_384:augreg_in21k_ft_in1k	nan
376	cvt_cvt-21-384-in1k_4_LucyV4	nan
377	resnet50_tutorial	nan
378	cvt_cvt-13-384-in1k_4_LucyV4	nan
379	alexnet_training_seed_08	nan
380	ReAlnet09	nan
381	yudixie_resnet18_translation_reg_0_240719	nan
382	yudixie_resnet18_distance_reg_0_240719	nan
383	ReAlnet10	nan
384	yudixie_resnet18_imagenet1kpret_0_240719	nan
385	mobilenet_v2_0_75_160	nan
386	mobilenet_v2_0_75_192	nan
387	alexnet_training_seed_10	nan
388	mobilenet_v2_0_5_224	nan
389	compact_resnet50_V4	nan
390	alexnet_training_seed_07	nan
391	compact_resnet50_IT	nan
392	resnet18_imagenet_full	nan
393	resnet50_imagenet_full	nan
394	resnet152_imagenet_full	nan
395	resnet18_ecoset_full	nan
396	alexnet_training_seed_01	nan
397	resnet34_ecoset_full	nan
398	resnet50_ecoset_full	nan
399	inception_v4
400	cvt_cvt-21-224-in1k_4_LucyV4	nan
401	resnet101_ecoset_full	nan
402	resnet50_imagenet_1_seed-0	nan
403	resnet152_ecoset_full	nan
404	resnet50_imagenet_10_seed-0	nan
405	yudixie_resnet18_distance_translation_0_240719	nan
406	yudixie_resnet18_translation_rotation_0_240719	nan
407	yudixie_resnet18_distance_rotation_0_240719	nan
408	yudixie_resnet18_category_class_0_240719	nan
409	yudixie_resnet18_random_0_240719	nan
410	yudixie_resnet18_cat_obj_class_all_latents_0_240719	nan
411	xception	nan
412	yudixie_resnet18_distance_translation_rotation_0_240719	nan
413	blt_vs	nan
414	deit_small_imagenet_1_seed-0	nan
415	nasnet_large
416	yudixie_resnet50_rotation_reg_0_240908	nan
417	yudixie_resnet50_distance_rotation_0_240908	nan
418	yudixie_resnet50_distance_translation_0_240908	nan
419	vgg_16	nan
420	pnasnet_large
421	compact_resnet50_robust_V4	nan
422	compact_vgg19_V4	nan
423	resnet_50_v1_spiking_l3	nan
424	compact_vgg19_IT	nan
425	nasnet_mobile	nan
426	mobilenet_v2_1_0_224	nan
427	yudixie_resnet50_translation_reg_0_240908	nan
428	mobilenet_v2_1_3_224	nan
429	mobilenet_v2_1_4_224	nan
430	compact_cornets_V4	nan
431	yudixie_resnet50_translation_rotation_0_240908	nan
432	convnext_xlarge:fb_in22k_ft_in1k	nan
433	resnet_50_v1	nan
434	vonegrcnn_62e_nobn
435	resnet_101_v2	nan
436	convnext_xxlarge:clip_laion2b_soup_ft_in1k	nan
437	swin_small_patch4_window7_224:ms_in22k_ft_in1k	nan
438	compact_cornets_IT	nan
439	artResNet18_1	nan
440	resnet_152_v2
441	resnet_101_v1
442	vit_base_patch16_clip_224:openai_ft_in12k_in1k	nan
443	vit_base_patch16_clip_224:openai_ft_in1k	nan
444	compact_resnet50_robust_IT	nan
445	vit_huge_patch14_clip_224:laion2b_ft_in12k_in1k	nan
446	vonegrcnn_47e
447	vonegrcnn_52e_full
448	BiT-S-R152x4
449	yudixie_resnet50_distance_translation_rotation_0_240908	nan
450	vit_huge_patch14_clip_336:laion2b_ft_in12k_in1k	nan
451	inception_v1	nan
452	vit_large_patch14_clip_224:laion2b_ft_in12k_in1k	nan
453	ReAlnet01	nan
454	vit_large_patch14_clip_224:openai_ft_in12k_in1k	nan
455	ReAlnet02	nan
456	densenet-121	nan
457	alexnet_training_seed_03	nan
458	vit_large_patch14_clip_224:openai_ft_in1k	nan
459	resnet50_primary_visual_cortex	nan
460	yudixie_resnet50_random_0_240908	nan
461	yudixie_resnet50_imagenet1kpret_0_240908	nan
462	yudixie_resnet50_cat_obj_class_all_latents_0_240908	nan
463	yudixie_resnet50_object_class_0_240908	nan
464	densenet-169	nan
465	convnext_base:clip_laiona_augreg_ft_in1k_384	nan
466	convnext_femto_ols:d1_in1k	nan

Benchmark bibtex

@Article{Kar2019,
                                                    author={Kar, Kohitij
                                                    and Kubilius, Jonas
                                                    and Schmidt, Kailyn
                                                    and Issa, Elias B.
                                                    and DiCarlo, James J.},
                                                    title={Evidence that recurrent circuits are critical to the ventral stream's execution of core object recognition behavior},
                                                    journal={Nature Neuroscience},
                                                    year={2019},
                                                    month={Jun},
                                                    day={01},
                                                    volume={22},
                                                    number={6},
                                                    pages={974-983},
                                                    abstract={Non-recurrent deep convolutional neural networks (CNNs) are currently the best at modeling core object recognition, a behavior that is supported by the densely recurrent primate ventral stream, culminating in the inferior temporal (IT) cortex. If recurrence is critical to this behavior, then primates should outperform feedforward-only deep CNNs for images that require additional recurrent processing beyond the feedforward IT response. Here we first used behavioral methods to discover hundreds of these `challenge' images. Second, using large-scale electrophysiology, we observed that behaviorally sufficient object identity solutions emerged {	extasciitilde}30{	hinspace}ms later in the IT cortex for challenge images compared with primate performance-matched `control' images. Third, these behaviorally critical late-phase IT response patterns were poorly predicted by feedforward deep CNN activations. Notably, very-deep CNNs and shallower recurrent CNNs better predicted these late IT responses, suggesting that there is a functional equivalence between additional nonlinear transformations and recurrence. Beyond arguing that recurrent circuits are critical for rapid object identification, our results provide strong constraints for future recurrent model development.},
                                                    issn={1546-1726},
                                                    doi={10.1038/s41593-019-0392-5},
                                                    url={https://doi.org/10.1038/s41593-019-0392-5}
                                                    }

Ceiling

0.79.

Note that scores are relative to this ceiling.

Data: Kar2019

1318 stimuli recordings from 424 sites in IT

Metric: ost