osmr / imgclsmob

TOP 1 ACCURACY TOP 5 ACCURACY
SPEED
MODEL CODE PAPER
ε-REPR
CODE PAPER
ε-REPR
PAPER
GLOBAL RANK
108-MENet-8x1
(g=3)
56.1% -- 79.2% -- #416
1.0-SqNxt-23
57.5% -- 80.9% -- #414
1.0-SqNxt-23v5
59.2% -- 82.2% -- #402
128-MENet-8x1
(g=4)
57.6% -- 80.4% -- #412
1.5-SqNxt-23
65.1% -- 86.5% -- #388
1.5-SqNxt-23v5
66.2% -- 87.0% -- #372
160-MENet-8x1
(g=8)
56.2% -- 79.2% -- #418
2.0-SqNxt-23
69.4% -- 89.0% -- #347
2.0-SqNxt-23v5
70.4% -- 89.3% -- #340
228-MENet-12x1
(g=3)
65.9% -- 86.8% -- #379
256-MENet-12x1
(g=4)
67.3% -- 87.5% -- #368
348-MENet-12x1
(g=3)
71.8% -- 90.4% -- #324
352-MENet-12x1
(g=8)
68.4% -- 88.0% -- #359
456-MENet-24x1
(g=3)
74.7% -- 92.0% -- #283
AlexNet
59.0% -- 81.8% -- 417.5 #406
AlexNet-b
58.4% -- 81.0% -- 418.1 #409
BAM-ResNet-50
76.9% -- 93.4% -- #224
BN-Inception
74.6% -- 92.3% -- 426.1 #279
BN-VGG-11
71.0% -- 90.4% -- 420.9 #323
BN-VGG-11b
70.4% -- 90.0% -- 415.6 #336
BN-VGG-13
72.2% -- 90.9% -- 406.0 #318
BN-VGG-13b
71.6% -- 90.4% -- 394.5 #324
BN-VGG-16
74.3% -- 92.2% -- 400.7 #288
BN-VGG-16b
72.8% -- 91.3% -- 394.4 #307
BN-VGG-19
75.9% -- 92.9% -- 395.8 #250
BN-VGG-19b
73.9% -- 91.6% -- 385.3 #296
CBAM-ResNet-50
77.6% -- 93.9% -- #190
CondenseNet-74
(C=G=4)
73.8% -- 91.7% -- #297
CondenseNet-74
(C=G=8)
71.1% -- 89.9% -- #333
DarkNet-53
78.3% -- 94.4% -- #179
DARTS
73.3% -- 91.3% -- #308
DenseNet-121
76.5% 76.4% 93.0% 93.3% #240
DenseNet-161
77.1% -- 93.6% -- #212
DenseNet-169
77.6% 77.9% 93.7% 94.1% #202
DenseNet-201
76.9% 78.5% 93.4% 94.5% #228
DiracNetV2-18
68.5% -- 88.3% -- #359
DiracNetV2-34
71.2% -- 90.1% -- #333
DLA-102
78.0% -- 94.0% -- #189
DLA-169
78.7% -- 94.3% -- #149
DLA-34
74.6% -- 92.1% -- #281
DLA-46-C
65.7% -- 86.8% -- #374
DLA-60
77.0% -- 93.3% -- #226
DLA-X-102
78.5% -- 94.2% -- #166
DLA-X2-102
79.5% -- 94.6% -- #109
DLA-X-46-C
66.7% -- 87.3% -- #368
DLA-X-60
78.2% -- 94.0% -- #182
DLA-X-60-C
69.0% -- 89.1% -- #353
DPN-131
79.5% -- 94.5% -- #107
DPN-68
76.8% -- 93.2% -- #235
DPN-98
79.2% -- 94.5% -- #124
DRN-C-26
75.1% -- 92.5% -- #265
DRN-C-42
77.1% -- 93.4% -- #222
DRN-C-58
78.3% -- 94.0% -- #185
DRN-D-105
79.4% -- 94.5% -- #112
DRN-D-22
74.2% -- 91.8% -- #290
DRN-D-38
76.2% -- 93.1% -- #241
DRN-D-54
78.8% -- 94.1% -- #149
EfficientNet-B0
75.2% 76.3% 92.5% 93.2% #261
EfficientNet-B0b
76.1% -- 93.0% -- #244
EfficientNet-B1
77.0% 78.8% 93.6% 94.4% #211
EfficientNet-B1b
78.4% -- 94.1% -- #177
EfficientNet-B2b
79.7% -- 94.7% -- 368.5 #101
EfficientNet-B3b
81.4% -- 95.6% -- 328.4 #55
EfficientNet-B4b
82.8% -- 96.2% -- 221.8 #41
EfficientNet-B5b
83.6% -- 96.7% -- 115.4 #27
EfficientNet-B6b
84.1% -- 96.9% -- 68.4 #22
EfficientNet-B7b
84.3% -- 96.9% -- 39.5 #14
ESPNetv2 x0.5
57.7% -- 79.8% -- #415
ESPNetv2 x1.0
66.1% -- 86.5% -- #380
ESPNetv2 x1.25
67.9% -- 87.8% -- #362
ESPNetv2 x1.5
69.2% -- 88.7% -- #352
ESPNetv2 x2.0
72.1% -- 90.4% -- #322
FBNet-Cb
75.1% -- 92.4% -- 407.7 #267
FD-MobileNet x0.25
44.1% -- 69.0% -- #424
FD-MobileNet x0.5
56.9% -- 79.8% -- #415
FD-MobileNet x0.75
61.6% -- 83.6% -- #396
FD-MobileNet x1.0
65.8% -- 86.6% -- #382
HRNetV2-W18
76.8% -- 93.4% -- 382.9 #236
HRNetV2-W30
78.2% -- 94.2% -- 377.8 #160
HRNetV2-W32
78.4% -- 94.2% -- 374.2 #165
HRNetV2-W40
78.9% -- 94.5% -- 354.0 #126
HRNetV2-W44
78.9% -- 94.4% -- 333.8 #143
HRNetV2-W48
79.3% -- 94.5% -- 315.6 #116
HRNetV2-W64
79.5% -- 94.7% -- 265.0 #106
HRNet-W18 Small V1
72.3% -- 90.7% -- 399.5 #315
HRNet-W18 Small V2
75.1% -- 92.4% -- 393.6 #267
IBN(b)-ResNet-50
76.4% -- 93.1% -- #242
IBN-DenseNet-121
75.5% -- 92.8% -- #255
IBN-DenseNet-169
76.8% -- 93.5% -- #216
IBN-ResNet-101
78.7% -- 94.4% -- #138
IBN-ResNet-50
77.2% -- 93.6% -- #217
IBN-ResNeXt-101
(32x4d)
79.1% -- 94.6% -- #115
IGCV3 x0.25
46.3% -- 71.3% -- #423
IGCV3 x0.5
60.2% -- 82.7% -- #399
IGCV3 x0.75
68.9% -- 88.6% -- #353
IGCV3 x1.0
72.1% -- 90.8% -- #314
InceptionResNetV
80.1% -- 95.1% -- #88
InceptionV3
78.9% -- 94.4% -- #147
InceptionV4
79.4% -- 94.7% -- #113
i-RevNet-301
74.0% -- 91.6% -- #294
MixNet-L
78.9% 78.9% 94.2% 94.2% 400.3 #168
MixNet-M
77.1% 77.0% 93.4% 93.3% 404.0 #223
MixNet-S
76.0% 75.8% 92.8% 92.8% 413.1 #250
MnasNet-A1
75.3% 75.2% 92.6% 92.5% 413.1 #258
MnasNet-B1
74.6% -- 92.2% -- 415.4 #280
MobileNetV2 x0.25
51.7% -- 75.5% -- #422
MobileNetV2 x0.5
64.0% -- 85.1% -- #393
MobileNetV2 x0.75
69.8% -- 89.2% -- #343
MobileNetV2 x1.0
73.0% -- 91.1% -- #308
MobileNetV3 L/224/1.0
75.1% -- 92.2% -- #264
MobileNet x0.25
53.7% -- 77.5% -- #420
MobileNet x0.5
65.8% -- 86.5% -- #384
MobileNet x0.75
69.9% -- 89.2% -- #341
MobileNet x1.0
73.4% -- 91.0% -- #311
74.3% -- 91.8% -- #287
81.9% -- 95.8% -- #51
PeleeNet
68.2% -- 88.5% -- #360
PNASNet-5-Large
82.1% -- 95.7% -- #44
PolyNet
80.9% 81.3% 95.5% 95.8% #62
PrepResNet-10
64.9% -- 85.8% -- #390
PreResNet-101
78.3% -- 94.1% -- #178
PreResNet-101b
79.0% -- 94.4% -- #129
PreResNet-12
66.1% -- 86.5% -- #373
PreResNet-14
67.4% -- 87.6% -- #367
PreResNet-152
79.1% -- 94.5% -- #131
PreResNet-152b
79.9% -- 94.8% -- #94
PreResNet-16
69.5% -- 88.9% -- #349
PreResNet-18
71.6% -- 90.3% -- #326
PreResNet-200b
78.7% -- 94.1% -- #158
PreResNet-26
73.7% -- 91.5% -- #299
PreResNet-269b
79.1% -- 94.2% -- #163
PreResNet-34
75.1% -- 92.3% -- #271
PreResNet-50
77.6% -- 93.5% -- #201
PreResNet-50b
77.5% -- 93.4% -- #208
PreResNet-BC-26b
74.5% -- 92.0% -- #285
PreResNet-BC-38b
77.1% -- 93.4% -- #221
ProxylessNAS CPU
75.3% -- 92.4% -- #265
ProxylessNAS GPU
75.2% -- 92.6% -- #259
ProxylessNAS Mob-14
76.7% -- 93.4% -- #227
ProxylessNAS Mobile
74.6% -- 92.2% -- #274
PyramidNet-101
(a=360)
78.0% -- 93.8% -- #192
ResNet-10
65.3% -- 85.6% -- #387
ResNet-101
78.1% 78.2% 93.8% 94.0% #196
ResNet-101b
79.4% -- 94.7% -- #106
ResNet-12
66.4% -- 86.7% -- #371
ResNet-14
67.5% -- 87.5% -- #365
ResNet-152
79.0% 78.6%
94.5% 94.3% #138
ResNet-152b
80.1% -- 95.0% -- #87
ResNet-16
69.5% -- 88.8% -- #348
ResNet-18
71.5% 72.1% 90.2% -- 412.9 #331
ResNet-26
73.7% -- 91.5% -- #300
ResNet-34
75.2% -- 92.2% -- #272
ResNet-50
77.7% 77.1%
93.7% 93.3%
#197
ResNet-50b
77.6% -- 93.6% -- #205
ResNet-BC-26b
74.9% -- 92.0% -- #272
ResNet-BC-38b
76.3% -- 93.0% -- #243
ResNeXt-101
(32x4d)
80.0% -- 94.8% -- #99
ResNeXt-101
(64x4d)
80.4% -- 94.9% -- #88
ResNeXt-14
(16x4d)
68.1% -- 87.5% -- #366
ResNeXt-14
(32x2d)
67.4% -- 87.2% -- #366
ResNeXt-14
(32x4d)
69.7% -- 88.5% -- #355
ResNeXt-26
(32x2d)
73.4% -- 91.1% -- #310
ResNeXt-26
(32x4d)
75.9% -- 92.5% -- #252
ResNeXt-50
(32x4d)
79.2% -- 94.4% -- #136
SENet-154
81.4% 82.7% 95.4% 96.2% #65
SENet-16
74.4% -- 91.8% -- #285
SENet-28
78.1% -- 94.0% -- #189
SE-PreResNet-10
66.0% -- 86.6% -- #377
SE-PreResNet-18
71.9% -- 90.4% -- #321
SE-PreResNet-BC-26b
76.8% -- 93.4% -- #225
SE-PreResNet-BC-38b
78.4% -- 94.2% -- #158
SE-ResNet-10
66.1% -- 86.3% -- #374
SE-ResNet-101
78.1% -- 94.1% -- #187
SE-ResNet-152
78.5% -- 94.2% -- #155
SE-ResNet-18
71.8% -- 90.4% -- #322
SE-ResNet-26
74.3% -- 91.8% -- #291
SE-ResNet-50
77.5% -- 93.6% -- #206
SE-ResNet-50b
79.2% -- 94.6% -- #122
SE-ResNet-BC-26b
76.4% -- 93.0% -- #245
SE-ResNet-BC-38b
78.4% -- 94.1% -- #179
SE-ResNeXt-101
(32x4d)
80.8% -- 95.2% -- 369.8 #73
SE-ResNeXt-101
(64x4d)
80.7% -- 95.2% -- 263.6 #71
SE-ResNeXt-50
(32x4d)
79.7% -- 94.8% -- 389.8 #98
ShuffleNetV2b x0.5
59.7% -- 81.8% -- #405
ShuffleNetV2b x1.0
69.4% -- 88.8% -- #351
ShuffleNetV2b x1.5
72.7% -- 90.9% -- #312
ShuffleNetV2b x2.0
74.4% -- 91.7% -- #295
ShuffleNetV2 x0.5
59.0% -- 81.4% -- #407
ShuffleNetV2 x1.0
68.6% -- 88.4% -- #357
ShuffleNetV2 x1.5
72.5% -- 90.6% -- #313
ShuffleNetV2 x2.0
74.1% -- 91.6% -- #298
ShuffleNet x0.25
(g=1)
37.6% -- 62.7% -- #426
ShuffleNet x0.25
(g=3)
38.3% -- 63.5% -- #425
ShuffleNet x0.5
(g=1)
53.4% -- 77.4% -- #421
ShuffleNet x0.5
(g=3)
55.8% -- 79.2% -- #419
ShuffleNet x0.75
(g=1)
60.4% -- 82.9% -- #399
ShuffleNet x0.75
(g=3)
61.8% -- 83.5% -- #395
ShuffleNet x1.0
(g=1)
65.1% -- 86.1% -- #389
ShuffleNet x1.0
(g=2)
65.8% -- 86.4% -- #385
ShuffleNet x1.0
(g=3)
65.6% -- 86.5% -- #386
ShuffleNet x1.0
(g=4)
65.8% -- 86.6% -- #381
ShuffleNet x1.0
(g=8)
65.9% -- 86.6% -- #379
SPNASNet
74.1% -- 91.8% -- 412.2 #292
SqueezeNet v1.0
60.7% -- 82.3% -- #400
SqueezeNet v1.1
60.7% -- 82.3% -- #398
SqueezeResNet v1.0
60.2% -- 81.9% -- #403
SqueezeResNet v1.1
59.9% -- 81.8% -- #404
VGG-11
70.1% 70.4% 89.6% 89.6% 411.2 #339
VGG-13
71.2% 71.3% 90.2% 90.1% 406.1 #332
VGG-16
73.0% 74.4% 91.3% 91.9% 396.4 #309
VGG-19
74.3% 74.5% 92.1% 92.0% 338.8 #289
WRN-50-2
77.5% -- 93.6% -- #210
Xception
79.0% 79.0% 94.5% 94.5% #137
ZFNet
60.2% 64.0% 82.7% 85.3% 428.2 #402
ZFNet-b
63.6% -- 85.1% -- 426.3 #392
See Full Build Details +get badge code
[![SotaBench](https://img.shields.io/endpoint.svg?url=https://sotabench.com/api/v0/badge/gh/osmr/imgclsmob)](https://sotabench.com/user/osemery/repos/osmr/imgclsmob)

How the Repository is Evaluated

The full sotabench.py file - source
from torchbench.image_classification import ImageNet
from pytorch.pytorchcv.model_provider import trained_model_metainfo_list
from pytorch.pytorchcv.model_provider import get_model as ptcv_get_model
import torchvision.transforms as transforms
import torch
import math
# import os

for model_metainfo in trained_model_metainfo_list:
    net = ptcv_get_model(model_metainfo[0], pretrained=True)
    input_image_size = model_metainfo[3]
    resize_inv_factor = model_metainfo[4]
    batch_size = model_metainfo[5]
    model_description = model_metainfo[6]
    assert (not hasattr(net, "in_size")) or (input_image_size == net.in_size[0])
    ImageNet.benchmark(
        model=net,
        model_description=model_description,
        paper_model_name=model_metainfo[1],
        paper_arxiv_id=model_metainfo[2],
        input_transform=transforms.Compose([
            transforms.Resize(int(math.ceil(float(input_image_size) / resize_inv_factor))),
            transforms.CenterCrop(input_image_size),
            transforms.ToTensor(),
            transforms.Normalize(
                mean=[0.485, 0.456, 0.406],
                std=[0.229, 0.224, 0.225]),
        ]),
        batch_size=batch_size,
        num_gpu=1,
        # data_root=os.path.join("..", "imgclsmob_data", "imagenet")
    )
    torch.cuda.empty_cache()
STATUS
BUILD
COMMIT MESSAGE
RUN TIME
Update PIP-packages (gluoncv2: 0.0.53, pytorchcv: 0.0.53, chaine…
osmr   72e817a  ·  2 days ago
0h:45m:27s
Remove unused script
osmr   a26629a  ·  3 days ago
0h:42m:11s
Update info about HRNet models
osmr   e93203e  ·  3 days ago
1h:06m:06s
1) Working on HRNet, 2) Some experiments with TF2
osmr   93610a1  ·  4 days ago
0h:42m:13s
Experiments with HRNet, 2
osmr   ff7423d  ·  5 days ago
0h:41m:11s
Experiments with HRNet
osmr   71920cd  ·  6 days ago
0h:43m:37s
Update info about VGG-19 model
osmr   8d635f0  ·  Nov 29 2019
0h:45m:31s
Working on TF2, 3
osmr   a2b859e  ·  Nov 26 2019
0h:41m:43s
Remove the previous version of MnasNet architecture
osmr   719fe9c  ·  Nov 24 2019
0h:41m:35s
Update PIP-packages (gluoncv2: 0.0.52, pytorchcv: 0.0.52, chaine…
osmr   df5b285  ·  Nov 24 2019
unknown
Update info about MnasNet models
osmr   1005d0d  ·  Nov 24 2019
0h:44m:38s
Working on MnasNet for TF1
osmr   b856755  (+4 commits )  ·  Nov 23 2019
0h:36m:54s
Update info about SE-ResNeXt-50/101 models
osmr   b6a154f  ·  Nov 20 2019
0h:50m:33s
Working on ResNeXt models
osmr   dbb02e9  ·  Nov 20 2019
0h:30m:34s
Working on TF2, 2
osmr   e73091e  ·  Nov 20 2019
0h:40m:51s
Working on TF2
osmr   b665d52  ·  Nov 18 2019
0h:41m:28s
Update PIP-packages (kerascv: 0.0.38)
osmr   f8e7fbb  ·  Nov 16 2019
0h:40m:29s
Fix a bug in EffNet for KE (fix #40)
osmr   92a0c62  ·  Nov 16 2019
0h:42m:43s
Prep for workin on MnasNet
osmr   fb220bf  ·  Nov 10 2019
0h:41m:15s
Update PIP-packages (gluoncv2: 0.0.51, pytorchcv: 0.0.51, chaine…
osmr   458c23b  ·  Nov 10 2019
0h:42m:28s
Cosmetic refactoring
osmr   abdece4  ·  Nov 10 2019
0h:41m:18s
Update info about SPNASNet model
osmr   2a0143a  ·  Nov 06 2019
0h:46m:53s
Update info about FBNet-Cb model
osmr   51d2d15  ·  Nov 06 2019
unknown
Update info about MixNet-L model
osmr   e40d1f7  ·  Nov 06 2019
0h:12m:35s
Experimenting with Sotabench, 6
osmr   ceb2e0f  ·  Nov 04 2019
1h:29m:18s
Experimenting with Sotabench, 5
osmr   b55ef4e  ·  Nov 04 2019
0h:23m:05s
Experimenting with Sotabench, 4
osmr 5cfd753    abb0a9e  ·  Nov 04 2019
1h:26m:57s
Experimenting with Sotabench, 2
osmr   6baac27  ·  Nov 02 2019
0h:42m:43s
Update info about MixNet-M model
osmr   cc191ac  ·  Nov 02 2019
6h:34m:35s
Experimenting with Sotabench
osmr   a7c8968  ·  Oct 29 2019
0h:52m:21s
0h:15m:25s
unknown
unknown
Update README
osmr   dcf351b  ·  Oct 29 2019
0h:04m:34s
Update info about MixNet-S model
osmr   6cc5e11  ·  Oct 29 2019
unknown
Add sotabench
osmr   d9e9877  ·  Oct 29 2019
unknown
unknown