Hugging Face's logo

timm documentation

ResNet

You are viewing main version, which requires installation from source. If you'd like regular pip install, checkout the latest stable version (v1.0.31).
Hugging Face's logo
Join the Hugging Face community

and get access to the augmented documentation experience

to get started

ResNet

카테고리: CNN · 코드: timm/models/resnet.py · 아키텍처: 81 · 프리트레이닝 가중치: 150 (ImageNet-1k 결과 포함 144)

Residual networks(ResNet) 및 동일한 ResNet 클래스를 기반으로 구축된 변형: ResNeXt 그룹 합성곱, Squeeze-and-Excitation(SE) 및 ECA 채널 어텐션, 딥/티어드 스템 및 평균 풀링 다운샘플링(“c”, “d”, “s”, “t” 변형), 안티앨리어스 다운샘플링, Wide ResNet 및 ResNet-RS 구성.

논문

pretrained weights에서 참조됨:

사용

사전 학습된 가중치 로드, 이미지 전처리 및 파인튜닝에 대해서는 Quickstart를, 모델을 백본으로 사용하는 방법에 대해서는 Feature Extraction을 참고하세요.

>>> import timm
>>> model = timm.create_model('resnet50.a1_in1k', pretrained=True)
>>> data_config = timm.data.resolve_model_data_config(model)

모델 API 지원

기능ResNet
features_only=True feature map✓ (strides 2, 4, 8, 16, 32)
forward_intermediates()✓
prune_intermediate_layers()✓
입력 크기 무관, 변경 불필요✓
set_input_size()–
dynamic_img_size=True–
set_grad_checkpointing()✓
레이어별 LR 감쇠(group_matcher())✓
reset_classifier()✓

✓ 지원됨, ✓ (정사각형) 정사각형 크기만 작동, ✗ 미지원, – 필요 없음(모델이 이미 모든 입력 크기를 허용함). 무작위로 초기화된 모델(ResNet의 resnet10t, 176)에서 각 기능을 실행하여 확인했습니다. 같은 클래스의 다른 아키텍처도 동일한 동작을 하는 것으로 예상됩니다. 자세한 내용은 Model API 페이지를 참조하세요.

아키텍처

아키텍처매개변수 (M)GMAC입력 크기사전 학습된 가중치
ecaresnet26t16.013.352561
ecaresnet50d25.584.352241
ecaresnet50d_pruned19.942.532241
ecaresnet50t25.575.642564
ecaresnet101d44.578.082241
ecaresnet101d_pruned24.883.482241
ecaresnet200d64.6920.002560
ecaresnet269d102.0941.533201
ecaresnetlight30.164.112241
ecaresnext26t_32x4d15.412.702240
ecaresnext50t_32x4d25.05-2240
resnet10t5.440.701761
resnet14t10.081.071761
resnet1811.691.822247
resnet18d11.712.062242
resnet2616.002.362241
resnet26d16.012.602241
resnet26t16.013.352561
resnet3421.803.672247
resnet34d21.823.912241
resnet5025.564.1122419
resnet50_gn25.564.142241
resnet50c25.584.352241
resnet50d25.584.352246
resnet50s25.685.472241
resnet50t25.574.322240
resnet10144.557.832247
resnet101c44.578.082241
resnet101d44.5710.552562
resnet101s44.679.192241
resnet15260.1911.562247
resnet152c60.2111.802241
resnet152d60.2115.412562
resnet152s60.3212.922241
resnet20064.6715.072240
resnet200d64.6920.002561
resnetaa34d21.824.432240
resnetaa5025.565.152241
resnetaa50d25.585.392243
resnetaa101d44.579.122242
resnetblur1811.692.342240
resnetblur5025.565.162241
resnetblur50d25.585.402240
resnetblur101d44.579.122240
resnetrs5035.692.291601
resnetrs10163.626.041921
resnetrs15286.6215.592561
resnetrs20093.2120.182561
resnetrs270129.8627.062561
resnetrs350163.9643.672881
resnetrs420191.8964.203201
resnext50_32x4d25.034.2622410
resnext50d_32x4d25.054.502241
resnext101_32x4d44.188.012243
resnext101_32x8d88.7916.482245
resnext101_32x16d194.0336.272243
resnext101_32x32d468.5387.292241
resnext101_64x4d83.4615.522243
senet154115.0920.772241
seresnet1811.781.822240
seresnet3421.963.672240
seresnet5028.094.112244
seresnet50t28.104.322240
seresnet10149.337.842240
seresnet15266.8211.572240
seresnet152d66.8415.422561
seresnet200d71.8620.012560
seresnet269d113.6726.592560
seresnetaa50d28.115.402240
seresnext26d_32x4d16.812.732241
seresnext26t_32x4d16.812.702241
seresnext50_32x4d27.564.262242
seresnext101_32x4d48.968.022241
seresnext101_32x8d93.5716.482241
seresnext101_64x4d88.2315.532241
seresnext101d_32x8d93.5916.722241
seresnextaa101d_32x8d93.5928.512884
seresnextaa201d_32x8d149.39101.113842
test_resnet0.470.101601
wide_resnet50_268.8811.432243
wide_resnet101_2126.8922.802242

GMACs와 매개변수 수는 각 아키텍처의 기본 입력 크기 및 분류기에 대한 값이며, 벤치마크 결과에서 가져왔습니다.

사전 학습된 가중치

Top-1은 results-imagenet.csv에 나열된 평가 크기에서 측정된 ImageNet-1k 검증 정확도입니다. - 가중치에 ImageNet-1k 결과가 없는 경우(예: 사전 학습 또는 이미지-텍스트 가중치)에는 표시하지 않습니다. Classes는 사전 학습된 분류기의 크기를 나타내며, 분류기가 없는 가중치의 경우 none입니다.

Weights는 소스에 따라 그룹화됩니다. 사전 학습 설정에 명시된 논문 또는 출처, 그렇지 않으면 timm가 아닌 Hub 조직, 그렇지 않으면 사전 학습 태그의 첫 번째 부분(architecture.tag)입니다. 단일 weight만 있는 소스는 other에 나열됩니다.

ResNet의 반격

Pretrained configs는 ResNet strikes back: An improved training procedure in timm를 참조합니다.

가중치입력 크기Top-1(%)평가 크기클래스라이선스Hub
ecaresnet50t.a1_in1k224 (테스트 288)82.1102881000apache-2.0timm
ecaresnet50t.a2_in1k224 (테스트 288)81.6822881000apache-2.0timm
ecaresnet50t.a3_in1k160 (테스트 224)79.5482241000apache-2.0timm
resnet18.a1_in1k224 (테스트 288)73.1642881000apache-2.0timm
resnet18.a2_in1k224 (테스트 288)72.3642881000apache-2.0timm
resnet18.a3_in1k160 (테스트 224)68.2502241000apache-2.0timm
resnet18d.ra4_e3600_r224_in1k224 (테스트 288)76.0122881000apache-2.0timm
resnet34.a1_in1k224 (테스트 288)77.9302881000apache-2.0timm
resnet34.a2_in1k224 (테스트 288)77.1502881000apache-2.0timm
resnet34.a3_in1k160 (테스트 224)72.9962241000apache-2.0timm
resnet34.ra4_e3600_r224_in1k224 (테스트 288)78.9722881000apache-2.0timm
resnet50.a1_in1k224 (테스트 288)81.2402881000apache-2.0timm
resnet50.a1h_in1k176 (테스트 224)80.6542241000apache-2.0timm
resnet50.a2_in1k224 (테스트 288)80.7742881000apache-2.0timm
resnet50.a3_in1k160 (테스트 224)78.0522241000apache-2.0timm
resnet50.b1k_in1k224 (테스트 288)80.7082881000apache-2.0timm
resnet50.b2k_in1k224 (테스트 288)80.4662881000apache-2.0timm
resnet50.c1_in1k224 (테스트 288)80.9282881000apache-2.0timm
resnet50.c2_in1k224 (테스트 288)80.8622881000apache-2.0timm
resnet50.d_in1k224 (테스트 288)81.0122881000apache-2.0timm
resnet50d.a1_in1k224 (테스트 288)81.4662881000apache-2.0timm
resnet50d.a2_in1k224 (테스트 288)81.1642881000apache-2.0timm
resnet50d.a3_in1k160 (테스트 224)78.7422241000apache-2.0timm
resnet50d.ra4_e3600_r224_in1k224 (테스트 288)81.8402881000apache-2.0timm
resnet101.a1_in1k224 (테스트 288)82.3262881000apache-2.0timm
resnet101.a1h_in1k224 (테스트 288)82.7962881000apache-2.0timm
resnet101.a2_in1k224 (테스트 288)82.2182881000apache-2.0timm
resnet101.a3_in1k160 (테스트 224)79.8362241000apache-2.0timm
resnet152.a1_in1k224 (테스트 288)82.7422881000apache-2.0timm
resnet152.a1h_in1k224 (테스트 288)83.4942881000apache-2.0timm
resnet152.a2_in1k224 (테스트 288)82.6142881000apache-2.0timm
resnet152.a3_in1k160 (테스트 224)80.5482241000apache-2.0timm
resnetaa50.a1h_in1k224 (테스트 288)81.6182881000apache-2.0timm
resnext50_32x4d.a1_in1k224 (테스트 288)81.4802881000apache-2.0timm
resnext50_32x4d.a1h_in1k224 (테스트 288)82.0162881000apache-2.0timm
resnext50_32x4d.a2_in1k224 (테스트 288)81.3042881000apache-2.0timm
resnext50_32x4d.a3_in1k160 (테스트 224)79.2582241000apache-2.0timm
resnext101_64x4d.c1_in1k224 (테스트 288)83.1582881000apache-2.0timm
seresnet50.a1_in1k224 (테스트 288)81.1142881000apache-2.0timm
seresnet50.a2_in1k224 (테스트 288)81.0942881000apache-2.0timm
seresnet50.a3_in1k160 (테스트 224)77.0362241000apache-2.0timm
seresnext101_32x8d.ah_in1k224 (테스트 288)84.1922881000apache-2.0timm
seresnext101d_32x8d.ah_in1k224 (테스트 288)84.3802881000apache-2.0timm
seresnextaa101d_32x8d.ah_in1k224 (테스트 288)84.5802881000apache-2.0timm

huggingface/pytorch-image-models

출처: github.com/huggingface/pytorch-image-models

가중치입력 크기Top-1(%)평가 크기클래스라이선스Hub
ecaresnet26t.ra2_in1k256 (테스트 320)79.9003201000apache-2.0timm
ecaresnet269d.ra2_in1k320 (테스트 352)84.9683521000apache-2.0timm
resnet10t.c3_in1k176 (테스트 224)68.3542241000apache-2.0timm
resnet14t.c3_in1k176 (테스트 224)72.2422241000apache-2.0timm
resnet18d.ra2_in1k224 (테스트 288)73.7902881000apache-2.0timm
resnet26.bt_in1k224 (테스트 288)76.3842881000apache-2.0timm
resnet26d.bt_in1k224 (테스트 288)77.4222881000apache-2.0timm
resnet26t.ra2_in1k256 (테스트 320)78.3303201000apache-2.0timm
resnet34.bt_in1k224 (테스트 288)76.5062881000apache-2.0timm
resnet34d.ra2_in1k224 (테스트 288)78.4402881000apache-2.0timm
resnet50.bt_in1k224 (테스트 288)79.6522881000apache-2.0timm
resnet50.ra_in1k224 (테스트 288)79.8642881000apache-2.0timm
resnet50.ram_in1k224 (테스트 288)79.9822881000apache-2.0timm
resnet50_gn.a1h_in1k224 (테스트 288)81.2162881000apache-2.0timm
resnet50d.ra2_in1k224 (테스트 288)81.3542881000apache-2.0timm
resnet101d.ra2_in1k256 (테스트 320)83.0203201000apache-2.0timm
resnet152d.ra2_in1k256 (테스트 320)83.6843201000apache-2.0timm
resnet200d.ra2_in1k256 (테스트 320)83.9763201000apache-2.0timm
resnetaa50d.d_in12k224 (테스트 288)--11821apache-2.0timm
resnetaa50d.sw_in12k224 (테스트 288)--11821apache-2.0timm
resnetaa50d.sw_in12k_ft_in1k224 (테스트 288)82.6562881000apache-2.0timm
resnetaa101d.sw_in12k224 (테스트 288)--11821apache-2.0timm
resnetaa101d.sw_in12k_ft_in1k224 (테스트 288)84.1142881000apache-2.0timm
resnetblur50.bt_in1k224 (테스트 288)80.2482881000apache-2.0timm
resnext50_32x4d.ra_in1k224 (테스트 288)80.7202881000apache-2.0timm
resnext50d_32x4d.bt_in1k224 (테스트 288)80.6762881000apache-2.0timm
seresnet50.ra2_in1k224 (테스트 288)81.2882881000apache-2.0timm
seresnet152d.ra2_in1k256 (테스트 320)84.3443201000apache-2.0timm
seresnext26d_32x4d.bt_in1k224 (테스트 288)78.8162881000apache-2.0timm
seresnext26t_32x4d.bt_in1k224 (테스트 288)78.7362881000apache-2.0timm
seresnext50_32x4d.racm_in1k224 (테스트 288)82.1862881000apache-2.0timm
seresnextaa101d_32x8d.sw_in12k224 (테스트 288)--11821apache-2.0timm
seresnextaa101d_32x8d.sw_in12k_ft_in1k224 (테스트 288)86.5242881000apache-2.0timm
seresnextaa101d_32x8d.sw_in12k_ft_in1k_288288 (테스트 320)86.7223201000apache-2.0timm
wide_resnet50_2.racm_in1k224 (테스트 288)82.2602881000apache-2.0timm

cv.gluon.ai

출처: cv.gluon.ai/model_zoo/classification.html

가중치입력 크기Top-1(%)평가 크기클래스라이선스Hub
resnet18.gluon_in1k22470.8442241000apache-2.0timm
resnet34.gluon_in1k22474.5742241000apache-2.0timm
resnet50.gluon_in1k22477.5802241000apache-2.0timm
resnet50c.gluon_in1k22478.0162241000apache-2.0timm
resnet50d.gluon_in1k22479.0802241000apache-2.0timm
resnet50s.gluon_in1k22478.7102241000apache-2.0timm
resnet101.gluon_in1k22479.3062241000apache-2.0timm
resnet101c.gluon_in1k22479.5422241000apache-2.0timm
resnet101d.gluon_in1k22480.4662241000apache-2.0timm
resnet101s.gluon_in1k22480.2782241000apache-2.0timm
resnet152.gluon_in1k22479.7122241000apache-2.0timm
resnet152c.gluon_in1k22479.9042241000apache-2.0timm
resnet152d.gluon_in1k22480.4762241000apache-2.0timm
resnet152s.gluon_in1k22481.0482241000apache-2.0timm
resnext50_32x4d.gluon_in1k22479.3542241000apache-2.0timm
resnext101_32x4d.gluon_in1k22480.3362241000apache-2.0timm
resnext101_64x4d.gluon_in1k22480.6402241000apache-2.0timm
senet154.gluon_in1k22481.2582241000apache-2.0timm
seresnext50_32x4d.gluon_in1k22479.9162241000apache-2.0timm
seresnext101_32x4d.gluon_in1k22480.8862241000apache-2.0timm
seresnext101_64x4d.gluon_in1k22480.8842241000apache-2.0timm

pytorch/vision

출처: github.com/pytorch/vision

가중치입력 크기Top-1(%)평가 크기클래스라이선스Hub
resnet18.tv_in1k22469.7522241000bsd-3-clausetimm
resnet34.tv_in1k22473.2982241000bsd-3-clausetimm
resnet50.tv2_in1k176 (테스트 224)80.8462241000bsd-3-clausetimm
resnet50.tv_in1k22476.1562241000bsd-3-clausetimm
resnet101.tv2_in1k176 (테스트 224)81.9062241000bsd-3-clausetimm
resnet101.tv_in1k22477.4042241000bsd-3-clausetimm
resnet152.tv2_in1k176 (테스트 224)82.2942241000bsd-3-clausetimm
resnet152.tv_in1k22478.3302241000bsd-3-clausetimm
resnext50_32x4d.tv2_in1k176 (테스트 224)81.1902241000bsd-3-clausetimm
resnext50_32x4d.tv_in1k22477.6322241000bsd-3-clausetimm
resnext101_32x8d.tv2_in1k176 (테스트 224)82.8322241000bsd-3-clausetimm
resnext101_32x8d.tv_in1k22479.3082241000bsd-3-clausetimm
resnext101_64x4d.tv_in1k22482.9782241000bsd-3-clausetimm
wide_resnet50_2.tv2_in1k176 (테스트 224)81.6122241000bsd-3-clausetimm
wide_resnet50_2.tv_in1k22478.4862241000bsd-3-clausetimm
wide_resnet101_2.tv2_in1k176 (테스트 224)82.5142241000bsd-3-clausetimm
wide_resnet101_2.tv_in1k22478.8622241000bsd-3-clausetimm

facebookresearch/semi-supervised-ImageNet1K-models

출처: github.com/facebookresearch/semi-supervised-ImageNet1K-models

가중치입력 크기Top-1(%)평가 크기클래스라이선스Hub
resnet18.fb_ssl_yfcc100m_ft_in1k22472.6362241000cc-by-nc-4.0timm
resnet18.fb_swsl_ig1b_ft_in1k22473.2862241000cc-by-nc-4.0timm
resnet50.fb_ssl_yfcc100m_ft_in1k22479.2522241000cc-by-nc-4.0timm
resnet50.fb_swsl_ig1b_ft_in1k22481.1462241000cc-by-nc-4.0timm
resnext50_32x4d.fb_ssl_yfcc100m_ft_in1k22480.3162241000cc-by-nc-4.0timm
resnext50_32x4d.fb_swsl_ig1b_ft_in1k22482.1622241000cc-by-nc-4.0timm
resnext101_32x4d.fb_ssl_yfcc100m_ft_in1k22480.9362241000cc-by-nc-4.0timm
resnext101_32x4d.fb_swsl_ig1b_ft_in1k22483.2682241000cc-by-nc-4.0timm
resnext101_32x8d.fb_ssl_yfcc100m_ft_in1k22481.6202241000cc-by-nc-4.0timm
resnext101_32x8d.fb_swsl_ig1b_ft_in1k22484.2762241000cc-by-nc-4.0timm
resnext101_32x16d.fb_ssl_yfcc100m_ft_in1k22481.8562241000cc-by-nc-4.0timm
resnext101_32x16d.fb_swsl_ig1b_ft_in1k22483.3402241000cc-by-nc-4.0timm

tf 태그

가중치입력 크기Top-1(%)평가 크기클래스라이선스Hub
resnetrs50.tf_in1k160 (테스트 224)79.9342241000apache-2.0timm
resnetrs101.tf_in1k192 (테스트 288)82.2882881000apache-2.0timm
resnetrs152.tf_in1k256 (테스트 320)83.7103201000apache-2.0timm
resnetrs200.tf_in1k256 (테스트 320)84.4483201000apache-2.0timm
resnetrs270.tf_in1k256 (테스트 352)84.4383521000apache-2.0timm
resnetrs350.tf_in1k288 (테스트 384)84.7103841000apache-2.0timm
resnetrs420.tf_in1k320 (테스트 416)85.0104161000apache-2.0timm

miil 태그

가중치입력 크기Top-1(%)평가 크기클래스라이선스Hub
ecaresnet50d.miil_in1k224 (테스트 288)81.6482881000apache-2.0timm
ecaresnet50d_pruned.miil_in1k224 (테스트 288)80.8082881000apache-2.0timm
ecaresnet101d.miil_in1k224 (테스트 288)83.0062881000apache-2.0timm
ecaresnet101d_pruned.miil_in1k224 (테스트 288)82.0162881000apache-2.0timm
ecaresnetlight.miil_in1k224 (테스트 288)81.4302881000apache-2.0timm

facebookresearch/WSL-Images

출처: github.com/facebookresearch/WSL-Images

가중치입력 크기Top-1(%)평가 크기클래스라이선스Hub
resnext101_32x8d.fb_wsl_ig1b_ft_in1k22482.7202241000cc-by-nc-4.0timm
resnext101_32x16d.fb_wsl_ig1b_ft_in1k22484.1662241000cc-by-nc-4.0timm
resnext101_32x32d.fb_wsl_ig1b_ft_in1k22485.0902241000cc-by-nc-4.0timm

sw 태그

가중치입력 크기Top-1(%)평가 크기클래스라이선스Hub
seresnextaa201d_32x8d.sw_in12k320 (테스트 384)--11821apache-2.0timm
seresnextaa201d_32x8d.sw_in12k_ft_in1k_38438487.3043841000apache-2.0timm

기타

가중치입력 크기Top-1(%)평가 크기클래스라이선스Hub
ecaresnet50t.ra2_in1k256 (테스트 320)82.3703201000apache-2.0timm
resnet50.am_in1k22479.0042241000apache-2.0timm
resnet50.bamboo224--없음-timm
test_resnet.r160_in1k16041.6021601000apache-2.0timm
Update on GitHub

←ResNeSt ResNetV2 (BiT)→