Scores on benchmarks
Model rank shown below is with respect to all public models.| .075 |
average_vision
rank 422
81 benchmarks |
|
| .015 |
neural_vision
rank 457
38 benchmarks |
|
| .059 |
V1
rank 443
24 benchmarks |
|
| .178 |
Coggan2024_fMRI.V1-rdm
v1
rank 29
|
|
|
||
| .135 |
behavior_vision
rank 219
43 benchmarks |
|
| .098 |
Geirhos2021-error_consistency
[reference]
rank 222
17 benchmarks |
|
| .179 |
Geirhos2021colour-error_consistency
v1
[reference]
rank 189
|
|
|
||
| .079 |
Geirhos2021contrast-error_consistency
v1
[reference]
rank 229
|
|
|
||
| .160 |
Geirhos2021cueconflict-error_consistency
v1
[reference]
rank 158
|
|
|
||
| .084 |
Geirhos2021edge-error_consistency
v1
[reference]
rank 131
|
|
|
||
| .077 |
Geirhos2021eidolonI-error_consistency
v1
[reference]
rank 275
|
|
|
||
| .173 |
Geirhos2021eidolonII-error_consistency
v1
[reference]
rank 223
|
|
|
||
| .123 |
Geirhos2021eidolonIII-error_consistency
v1
[reference]
rank 240
|
|
|
||
| .218 |
Geirhos2021falsecolour-error_consistency
v1
[reference]
rank 169
|
|
|
||
| .048 |
Geirhos2021highpass-error_consistency
v1
[reference]
rank 169
|
|
|
||
| .090 |
Geirhos2021lowpass-error_consistency
v1
[reference]
rank 195
|
|
|
||
| .051 |
Geirhos2021phasescrambling-error_consistency
v1
[reference]
rank 218
|
|
|
||
| .013 |
Geirhos2021powerequalisation-error_consistency
v1
[reference]
rank 279
|
|
|
||
| .063 |
Geirhos2021rotation-error_consistency
v1
[reference]
rank 206
|
|
|
||
| .052 |
Geirhos2021silhouette-error_consistency
v1
[reference]
rank 289
|
|
|
||
| .044 |
Geirhos2021sketch-error_consistency
v1
[reference]
rank 222
|
|
|
||
| .159 |
Geirhos2021stylized-error_consistency
v1
[reference]
rank 184
|
|
|
||
| .061 |
Geirhos2021uniformnoise-error_consistency
v1
[reference]
rank 184
|
|
|
||
| .091 |
Baker2022
rank 153
3 benchmarks |
|
| .274 |
Baker2022fragmented-accuracy_delta
v1
[reference]
rank 115
|
|
|
||
| .000 |
Baker2022frankenstein-accuracy_delta
v1
[reference]
rank 145
|
|
|
||
| .000 |
Baker2022inverted-accuracy_delta
v1
[reference]
rank 55
|
|
|
||
| .146 |
BMD2024
rank 127
4 benchmarks |
|
| .166 |
BMD2024.dotted_1Behavioral-accuracy_distance
v1
rank 84
|
|
|
||
| .126 |
BMD2024.dotted_2Behavioral-accuracy_distance
v1
rank 109
|
|
|
||
| .134 |
BMD2024.texture_1Behavioral-accuracy_distance
v1
rank 127
|
|
|
||
| .157 |
BMD2024.texture_2Behavioral-accuracy_distance
v1
rank 118
|
|
|
||
| .233 |
Hebart2023-match
v1
rank 141
|
|
|
||
| .417 |
Maniquet2024
rank 158
2 benchmarks |
|
| .164 |
Maniquet2024-confusion_similarity
v1
[reference]
rank 187
|
|
|
||
| .671 |
Maniquet2024-tasks_consistency
v1
[reference]
rank 69
|
|
|
||
| .095 |
Coggan2024_behavior-ConditionWiseAccuracySimilarity
v1
rank 173
|
|
|
||
| .104 |
engineering_vision
rank 274
25 benchmarks |
|
| .520 |
Geirhos2021-top1
[reference]
rank 155
17 benchmarks |
|
| .948 |
Geirhos2021colour-top1
v1
[reference]
rank 152
|
|
|
||
| .855 |
Geirhos2021contrast-top1
v1
[reference]
rank 94
|
|
|
||
| .158 |
Geirhos2021cueconflict-top1
v1
[reference]
rank 238
|
|
|
||
| .194 |
Geirhos2021edge-top1
v1
[reference]
rank 205
|
|
|
||
| .497 |
Geirhos2021eidolonI-top1
v1
[reference]
rank 139
|
|
|
||
| .511 |
Geirhos2021eidolonII-top1
v1
[reference]
rank 142
|
|
|
||
| .498 |
Geirhos2021eidolonIII-top1
v1
[reference]
rank 160
|
|
|
||
| .929 |
Geirhos2021falsecolour-top1
v1
[reference]
rank 133
|
|
|
||
| .500 |
Geirhos2021highpass-top1
v1
[reference]
rank 66
|
|
|
||
| .419 |
Geirhos2021lowpass-top1
v1
[reference]
rank 133
|
|
|
||
| .539 |
Geirhos2021phasescrambling-top1
v1
[reference]
rank 186
|
|
|
||
| .598 |
Geirhos2021powerequalisation-top1
v1
[reference]
rank 182
|
|
|
||
| .674 |
Geirhos2021rotation-top1
v1
[reference]
rank 127
|
|
|
||
| .319 |
Geirhos2021silhouette-top1
v1
[reference]
rank 235
|
|
|
||
| .487 |
Geirhos2021sketch-top1
v1
[reference]
rank 212
|
|
|
||
| .329 |
Geirhos2021stylized-top1
v1
[reference]
rank 204
|
|
|
||
| .381 |
Geirhos2021uniformnoise-top1
v1
[reference]
rank 152
|
|
|
||
How to use
from brainscore_vision import load_model
model = load_model("resnet50_eMMCR_Vanilla")
model.start_task(...)
model.start_recording(...)
model.look_at(...)
Benchmarks bibtex
@inproceedings{santurkar2019computer,
title={Computer Vision with a Single (Robust) Classifier},
author={Shibani Santurkar and Dimitris Tsipras and Brandon Tran and Andrew Ilyas and Logan Engstrom and Aleksander Madry},
booktitle={ArXiv preprint arXiv:1906.09453},
year={2019}
}
@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}
}
@article{BAKER2022104913,
title = {Deep learning models fail to capture the configural nature of human shape perception},
journal = {iScience},
volume = {25},
number = {9},
pages = {104913},
year = {2022},
issn = {2589-0042},
doi = {https://doi.org/10.1016/j.isci.2022.104913},
url = {https://www.sciencedirect.com/science/article/pii/S2589004222011853},
author = {Nicholas Baker and James H. Elder},
keywords = {Biological sciences, Neuroscience, Sensory neuroscience},
abstract = {Summary
A hallmark of human object perception is sensitivity to the holistic configuration of the local shape features of an object. Deep convolutional neural networks (DCNNs) are currently the dominant models for object recognition processing in the visual cortex, but do they capture this configural sensitivity? To answer this question, we employed a dataset of animal silhouettes and created a variant of this dataset that disrupts the configuration of each object while preserving local features. While human performance was impacted by this manipulation, DCNN performance was not, indicating insensitivity to object configuration. Modifications to training and architecture to make networks more brain-like did not lead to configural processing, and none of the networks were able to accurately predict trial-by-trial human object judgements. We speculate that to match human configural sensitivity, networks must be trained to solve a broader range of object tasks beyond category recognition.}
}
@article {Maniquet2024.04.02.587669,
author = {Maniquet, Tim and de Beeck, Hans Op and Costantino, Andrea Ivan},
title = {Recurrent issues with deep neural network models of visual recognition},
elocation-id = {2024.04.02.587669},
year = {2024},
doi = {10.1101/2024.04.02.587669},
publisher = {Cold Spring Harbor Laboratory},
URL = {https://www.biorxiv.org/content/early/2024/04/10/2024.04.02.587669},
eprint = {https://www.biorxiv.org/content/early/2024/04/10/2024.04.02.587669.full.pdf},
journal = {bioRxiv}
}