Self-Supervised Image Classification 벤치마크

Self-Supervised Image Classification on ImageNet

288개 결과 · ⬇ CSV · JSON

Top 1 Accuracy

Rank	Model	Top 1 Accuracy	Top 5 Accuracy	Number of Params	Extra Training Data	Paper	Code	Year
101	SynCo (ResNet-50) 800ep	70.6%	89.8%	24M		SynCo: Synthetic Hard Negatives in Contrastive Learning for Better Unsupervised Visual Representations	giakoumoglou/synco	2024
101	CMC (ResNet-50 x2) (arxiv v5)	70.6%	89.7%	188M		Contrastive Multiview Coding	HobbitLong/PyContrast · HobbitLong/CMC · szq0214/Rethinking-Image-Mixture-for-Unsupervised-Learning · +5	2019
103	SimCLR (ResNet-50)	69.3%	89.0%	24M		A Simple Framework for Contrastive Learning of Visual Representations	tensorflow/models · google-research/simclr · lightly-ai/lightly · +93	2020
104	iGPT-XL (64x64, 3072 features)	68.7%	–	6800M		Generative Pretraining from Pixels	openai/image-gpt · EugenHotaj/pytorch-generative · teddykoker/image-gpt · +1	2020
105	MoCo (ResNet-50 4x)	68.6%	–	375M		Momentum Contrast for Unsupervised Visual Representation Learning	open-mmlab/mmdetection · KevinMusgrave/pytorch-metric-learning · KevinMusgrave/pytorch_metric_learning · +41	2019
106	AMDIM (large) (arxiv v2)	68.1%	–	626M		Learning Representations by Maximizing Mutual Information Across Views	philip-bachman/amdim-public · Alibaba-AAIG/SSL-FEW-SHOT · cfld/amdim	2019
107	MAE (ViT-B)	68.0%	–	80M		Masked Autoencoders Are Scalable Vision Learners	facebookresearch/mae · lightly-ai/lightly · open-mmlab/mmselfsup · +55	2021
108	SynCo (ResNet-50) 200ep	67.9%	88	24M		SynCo: Synthetic Hard Negatives in Contrastive Learning for Better Unsupervised Visual Representations	giakoumoglou/synco	2024
109	DINO (ResMLP-12)	67.5%	–	15M		ResMLP: Feedforward networks for image classification with data-efficient training	rwightman/pytorch-image-models · xmu-xiaoma666/External-Attention-pytorch · facebookresearch/deit · +16	2021
110	CMC (ResNet-50) (arxiv v5)	66.2%	87.0%	47M		Contrastive Multiview Coding	HobbitLong/PyContrast · HobbitLong/CMC · szq0214/Rethinking-Image-Mixture-for-Unsupervised-Learning · +5	2019
111	PCL (ResNet-50)	65.9%	–	25M		Prototypical Contrastive Learning of Unsupervised Representations	salesforce/PCL · salesforce/MoPro	2020
112	MoCo (ResNet-50 2x)	65.4%	–	94M		Momentum Contrast for Unsupervised Visual Representation Learning	open-mmlab/mmdetection · KevinMusgrave/pytorch-metric-learning · KevinMusgrave/pytorch_metric_learning · +41	2019
113	iGPT-L (48x48)	65.2%	–	1400M		Generative Pretraining from Pixels	openai/image-gpt · EugenHotaj/pytorch-generative · teddykoker/image-gpt · +1	2020
114	CMC (ResNet-101) (arxiv v3)	65.0%	86.0%	–		Contrastive Multiview Coding	HobbitLong/PyContrast · HobbitLong/CMC · szq0214/Rethinking-Image-Mixture-for-Unsupervised-Learning · +5	2019
115	CPC v2 (ResNet-50) (arxiv v2)	63.8%	85.3%	24M		Data-Efficient Image Recognition with Contrastive Predictive Coding	philip-bachman/amdim-public · mf1024/Contrastive-Predictive-Coding-for-Image-Recognition-in-PyTorch · SeonghoBaek/FrameSequencePrediction · +1	2019
115	MMCL (100 epoch, 256 batch size)	63.8%	–	–		Max-Margin Contrastive Learning	anshulbshah/MMCL	2021
117	PIRL	63.6%	–	24M		Self-Supervised Learning of Pretext-Invariant Representations	facebookresearch/vissl · HobbitLong/PyContrast · akwasigroch/Pretext-Invariant-Representations · +4	2019
118	AMDIM (small) (arxiv v2)	63.5%	–	194M		Learning Representations by Maximizing Mutual Information Across Views	philip-bachman/amdim-public · Alibaba-AAIG/SSL-FEW-SHOT · cfld/amdim	2019
119	SeLa (ResNet50) (arxiv 3)	61.5%	84.0%	24M		Self-labelling via simultaneous clustering and representation learning	yukimasano/self-label · mingu6/action_seg_ot · hsfzxjy/swavx · +2	2019
120	BigBiGAN (RevNet-50 ×4, BN+CReLU)	61.3%	81.9%	86M		Large Scale Adversarial Representation Learning	lukemelas/unsupervised-image-segmentation · LEGO999/BIgBiGAN · LEGO999/BigBiGAN-TensorFlow2.0 · +1	2019
121	CPC v2 (ResNet-161) (arxiv v1)	61.0%	83.0%	305M		Data-Efficient Image Recognition with Contrastive Predictive Coding	philip-bachman/amdim-public · mf1024/Contrastive-Predictive-Coding-for-Image-Recognition-in-PyTorch · SeonghoBaek/FrameSequencePrediction · +1	2019
122	BigBiGAN (RevNet-50 ×4)	60.8%	81.4%	86M		Large Scale Adversarial Representation Learning	lukemelas/unsupervised-image-segmentation · LEGO999/BIgBiGAN · LEGO999/BigBiGAN-TensorFlow2.0 · +1	2019
123	MoCo (ResNet-50)	60.6%	–	24M		Momentum Contrast for Unsupervised Visual Representation Learning	open-mmlab/mmdetection · KevinMusgrave/pytorch-metric-learning · KevinMusgrave/pytorch_metric_learning · +41	2019
124	iGPT-L (32x32)	60.3%	–	1400M		Generative Pretraining from Pixels	openai/image-gpt · EugenHotaj/pytorch-generative · teddykoker/image-gpt · +1	2020
125	AMDIM (arxiv v1)	60.2%	–	337M		Learning Representations by Maximizing Mutual Information Across Views	philip-bachman/amdim-public · Alibaba-AAIG/SSL-FEW-SHOT · cfld/amdim	2019
125	LocalAgg (ResNet-50)	60.2%	–	24M		Local Aggregation for Unsupervised Learning of Visual Embeddings	neuroailab/LocalAggregation-Pytorch	2019
127	CMC (ResNet-101)	60.1%	82.8%	44M		Contrastive Multiview Coding	HobbitLong/PyContrast · HobbitLong/CMC · szq0214/Rethinking-Image-Mixture-for-Unsupervised-Learning · +5	2019
128	BigBiGAN (ResNet-50, BN+CReLU)	56.6%	78.6%	24M		Large Scale Adversarial Representation Learning	lukemelas/unsupervised-image-segmentation · LEGO999/BIgBiGAN · LEGO999/BigBiGAN-TensorFlow2.0 · +1	2019
129	SeLa (ResNet50)	55.7%	79.5%	24M		Self-labelling via simultaneous clustering and representation learning	yukimasano/self-label · mingu6/action_seg_ot · hsfzxjy/swavx · +2	2019
130	Revisited Rotation (RevNet-50 ×4)	55.4%	77.9%	86M		Revisiting Self-Supervised Visual Representation Learning	philip-bachman/amdim-public · google/revisiting-self-supervised · virtualgraham/sc_patch · +3	2019
130	BigBiGAN (ResNet-50)	55.4%	77.4%	25M		Large Scale Adversarial Representation Learning	lukemelas/unsupervised-image-segmentation · LEGO999/BIgBiGAN · LEGO999/BigBiGAN-TensorFlow2.0 · +1	2019
132	Revisited Rel.Patch.Loc (ResNet50 ×2)	51.4%	74.0%	94M		Revisiting Self-Supervised Visual Representation Learning	philip-bachman/amdim-public · google/revisiting-self-supervised · virtualgraham/sc_patch · +3	2019
133	SeLa (AlexNet) (arxiv v3)	50.0%	–	61M		Self-labelling via simultaneous clustering and representation learning	yukimasano/self-label · mingu6/action_seg_ot · hsfzxjy/swavx · +2	2019
134	CPC (ResNet-101 V2)	48.7%	73.6%	44M		Representation Learning with Contrastive Predictive Coding	RElbers/info-nce-pytorch · davidtellez/contrastive-predictive-coding · jefflai108/Contrastive-Predictive-Coding-PyTorch · +25	2018
135	Revisited Exemplar (ResNet-50 ×3)	46.0%	68.8%	211M		Revisiting Self-Supervised Visual Representation Learning	philip-bachman/amdim-public · google/revisiting-self-supervised · virtualgraham/sc_patch · +3	2019
136	Revisited Jigsaw (ResNet50 ×2)	44.6%	68.0%	94M		Revisiting Self-Supervised Visual Representation Learning	philip-bachman/amdim-public · google/revisiting-self-supervised · virtualgraham/sc_patch · +3	2019
137	CMC (Alexnet/2)	42.6%	–	30M		Contrastive Multiview Coding	HobbitLong/PyContrast · HobbitLong/CMC · szq0214/Rethinking-Image-Mixture-for-Unsupervised-Learning · +5	2019
138	DeepCluster (AlexNet)	41.0	–	61M		Deep Clustering for Unsupervised Learning of Visual Features	facebookresearch/deepcluster · bdy9527/SDCN · 461054993/SDCN · +6	2018
139	Colorisation (improved) (ResNet-101)	39.6	62.5	44M		Multi-task Self-Supervised Visual Learning		2017
140	Rotation (AlexNet)	38.7	–	86M		Unsupervised Representation Learning by Predicting Image Rotations	facebookresearch/vissl · open-mmlab/mmselfsup · YyzHarry/imbalanced-semi-self · +17	2018
141	Split-Brain (AlexNet)	35.4%	–	61M		Split-Brain Autoencoders: Unsupervised Learning by Cross-Channel Prediction	richzhang/splitbrainauto · ysharma1126/Split-Brain-Autoencoder	2016
142	Counting (AlexNet)	34.3	–	61M		Representation Learning by Learning to Count	clvrai/representation-learning-by-learning-to-count · gitlimlab/Representation-Learning-by-Learning-to-Count	2017
143	Colorization (AlexNet)	32.6%	–	61M		Colorful Image Colorization	richzhang/colorization · baldassarreFe/deep-koalarization · demul/auto_colorization_project · +36	2016
144	Multi-task SSL (ResNet-101)	–	70.2	44M		Multi-task Self-Supervised Visual Learning		2017
145	DINOv2+reg (ViT-g/14)	87.1	–	1100M	✓	Vision Transformers Need Registers	rwightman/pytorch-image-models · facebookresearch/dinov2 · locuslab/massive-activations · +3	2023
146	DINOv2 (ViT-g/14 @448)	86.7%	–	1100M	✓	DINOv2: Learning Robust Visual Features without Supervision	huggingface/transformers · facebookresearch/dinov2 · roboflow/rf-detr · +23	2023
147	DINOv2 (ViT-g/14)	86.5%	–	1100M	✓	DINOv2: Learning Robust Visual Features without Supervision	huggingface/transformers · facebookresearch/dinov2 · roboflow/rf-detr · +23	2023
148	DINOv2 distilled (ViT-L/14)	86.3%	–	307M	✓	DINOv2: Learning Robust Visual Features without Supervision	huggingface/transformers · facebookresearch/dinov2 · roboflow/rf-detr · +23	2023
149	MIM-Refiner (D2V2-ViT-H/14)	84.7%	–	632M		MIM-Refiner: A Contrastive Learning Boost from Intermediate Pre-Trained Representations	ml-jku/MIM-Refiner · BenediktAlkin/vtab1k-pytorch	2024
150	MIM-Refiner (MAE-ViT-2B/14)	84.5%	–	1890M		MIM-Refiner: A Contrastive Learning Boost from Intermediate Pre-Trained Representations	ml-jku/MIM-Refiner · BenediktAlkin/vtab1k-pytorch	2024
150	DINOv2 distilled (ViT-B/14)	84.5%	–	85M	✓	DINOv2: Learning Robust Visual Features without Supervision	huggingface/transformers · facebookresearch/dinov2 · roboflow/rf-detr · +23	2023
152	MIM-Refiner (MAE-ViT-H/14	83.7%	–	632M		MIM-Refiner: A Contrastive Learning Boost from Intermediate Pre-Trained Representations	ml-jku/MIM-Refiner · BenediktAlkin/vtab1k-pytorch	2024
153	MIM-Refiner (D2V2-ViT-L/16)	83.5%	–	307M		MIM-Refiner: A Contrastive Learning Boost from Intermediate Pre-Trained Representations	ml-jku/MIM-Refiner · BenediktAlkin/vtab1k-pytorch	2024
154	MIM-Refiner (MAE-ViT-L/16)	82.8%	–	307M		MIM-Refiner: A Contrastive Learning Boost from Intermediate Pre-Trained Representations	ml-jku/MIM-Refiner · BenediktAlkin/vtab1k-pytorch	2024
155	iBOT (ViT-L/16) (IN22k)	82.3%	–	307M	✓	iBOT: Image BERT Pre-Training with Online Tokenizer	bytedance/ibot · birder/birder	2021
156	MAE-CT (ViT-H/16)	82.2%	–	632M		Contrastive Tuning: A Little Help to Make Masked Autoencoders Forget	ml-jku/mae-ct	2023
157	Mugs (VIT-L/16)	82.1%	–	307M		Mugs: A Multi-Granular Self-Supervised Learning Framework	sail-sg/mugs	2022
158	MAE-CT (ViT-L/16	81.5%	–	307M		Contrastive Tuning: A Little Help to Make Masked Autoencoders Forget	ml-jku/mae-ct	2023
159	EsViT (Swin-B)	81.3	95.5	87M		Efficient Self-supervised Vision Transformers for Representation Learning	microsoft/esvit	2021
159	iBOT (ViT-L/16)	81.3%	–	307M		iBOT: Image BERT Pre-Training with Online Tokenizer	bytedance/ibot · birder/birder	2021
161	DINOv2 distilled (ViT-S/14)	81.1%	–	21M	✓	DINOv2: Learning Robust Visual Features without Supervision	huggingface/transformers · facebookresearch/dinov2 · roboflow/rf-detr · +23	2023
162	MoCo v3 (ViT-BN-L/7)	81.0%	–	304M		An Empirical Study of Training Self-Supervised Vision Transformers	open-mmlab/mmselfsup · facebookresearch/moco-v3 · Westlake-AI/openmixup · +6	2021
163	EsViT(Swin-S)	80.8	–	49M		Efficient Self-supervised Vision Transformers for Representation Learning	microsoft/esvit	2021
164	MSN (ViT-L/7)	80.7%	–	306M		Masked Siamese Networks for Label-Efficient Learning	lightly-ai/lightly · facebookresearch/msn	2022
165	ReLICv2 (ResNet-200 x2)	80.6%	–	250M		Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?	google-deepmind/relicv2	2022
166	MR BarTwins (MR BarTwins)	80.4%	–	–		Masked Reconstruction Contrastive Learning with Information Bottleneck Principle		2022
167	DiGIT	80.3%	–	732M		Stabilize the Latent Space for Image Autoregressive Modeling: A Unified Perspective	DAMO-NLP-SG/DiGIT	2024
167	iBOT-vMF (ViT-B/16)	80.3%	–	85M		DINO as a von Mises-Fisher mixture model		2024
167	DINO (xcit_medium_24_p8)	80.3%	–	84M		Emerging Properties in Self-Supervised Vision Transformers	facebookresearch/dino · lightly-ai/lightly · facebookresearch/vissl · +29	2021
167	PGT (PGT-B w/ Flow)	80.3%	–	70M		Perceptual Group Tokenizer: Building Perception with Iterative Grouping		2023
171	DINO (ViT-B/8)	80.1%	–	80M		Emerging Properties in Self-Supervised Vision Transformers	facebookresearch/dino · lightly-ai/lightly · facebookresearch/vissl · +29	2021
172	SimCLRv2 (ResNet-152 x3, SK)	79.8%	94.9%	795M		Big Self-Supervised Models are Strong Semi-Supervised Learners	google-research/simclr · lightly-ai/lightly · sayakpaul/PAWS-TF · +6	2020
172	SEERv2	79.8%	–	10000M	✓	Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision	facebookresearch/vissl	2022
172	PercMAE (ViT-B, dVAE)	79.8%	–	80M	✓	Improving Visual Representation Learning through Perceptual Understanding	tractableai/perceptual-mae	2022
172	ReLICv2 (ResNet200)	79.8%	–	63M		Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?	google-deepmind/relicv2	2022
176	DINO (ViT-S/8)	79.7%	–	21M		Emerging Properties in Self-Supervised Vision Transformers	facebookresearch/dino · lightly-ai/lightly · facebookresearch/vissl · +29	2021
177	BYOL (ResNet-200 x2)	79.6%	94.8%	250M		Bootstrap your own latent: A new approach to self-supervised Learning	deepmind/deepmind-research · alibaba/EasyCV · lucidrains/byol-pytorch · +28	2020
178	ReLICv2 (ResNet-50 4x)	79.4%	–	375M		Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?	google-deepmind/relicv2	2022
179	ReLICv2 (ResNet152)	79.3%	–	58M		Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?	google-deepmind/relicv2	2022
179	BAM (CAFormer-M36)	79.3%	–	–		Unsupervised Representation Learning by Balanced Self Attention Matching	danielshalam/bam	2024
181	MoCo v3 (ViT-BN-H)	79.1%	–	700M		An Empirical Study of Training Self-Supervised Vision Transformers	open-mmlab/mmselfsup · facebookresearch/moco-v3 · Westlake-AI/openmixup · +6	2021
181	Unicom (ViT-B/16)	79.1%	–	80M	✓	Unicom: Universal and Compact Representation Learning for Image Retrieval	OML-Team/open-metric-learning · deepglint/unicom · RocketFlash/easy_metric_learning	2023
183	SMoG (ResNet-50 x4)	79.0%	94.4	375M		Unsupervised Visual Representation Learning by Synchronous Momentum Grouping	lightly-ai/lightly	2022
183	ReLICv2 (ResNet-50 x2)	79%	–	94M		Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?	google-deepmind/relicv2	2022
185	C-BYOL (ResNet-50 2x, 1000 epochs)	78.8%	94.5%	94M		Compressive Visual Representations	google-research/compressive-visual-representations	2021
185	DINO-vMF (ViT-B/16)	78.8%	–	85M		DINO as a von Mises-Fisher mixture model		2024
187	ReLICv2 (ResNet101)	78.7%	–	44M		Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?	google-deepmind/relicv2	2022
188	BYOL (ResNet-50 x4)	78.6%	94.2%	375M		Bootstrap your own latent: A new approach to self-supervised Learning	deepmind/deepmind-research · alibaba/EasyCV · lucidrains/byol-pytorch · +28	2020
189	SwAV (ResNet-50 x5)	78.5%	–	586M		Unsupervised Learning of Visual Features by Contrasting Cluster Assignments	open-mmlab/mmdetection · lightly-ai/lightly · facebookresearch/vissl · +15	2020
190	DINO (ViT-B/16)	78.2%	–	85M		Emerging Properties in Self-Supervised Vision Transformers	facebookresearch/dino · lightly-ai/lightly · facebookresearch/vissl · +29	2021
191	MoCo v3 (ViT-H)	78.1%	–	632M		An Empirical Study of Training Self-Supervised Vision Transformers	open-mmlab/mmselfsup · facebookresearch/moco-v3 · Westlake-AI/openmixup · +6	2021
191	PercMAE (ViT-B)	78.1%	–	80M		Improving Visual Representation Learning through Perceptual Understanding	tractableai/perceptual-mae	2022
191	BAM (ViT-B/16)	78.1%	–	80M		Unsupervised Representation Learning by Balanced Self Attention Matching	danielshalam/bam	2024
194	SMoG (ResNet-50 x2)	78.0%	93.9	94M		Unsupervised Visual Representation Learning by Synchronous Momentum Grouping	lightly-ai/lightly	2022
195	MoCo v3 (ViT-L)	77.6%	–	307M		An Empirical Study of Training Self-Supervised Vision Transformers	open-mmlab/mmselfsup · facebookresearch/moco-v3 · Westlake-AI/openmixup · +6	2021
196	SEER	77.5%	–	1300M	✓	Self-supervised Pretraining of Visual Features in the Wild	facebookresearch/vissl	2021
197	BYOL (ResNet-50 x2)	77.4%	93.6%	94M		Bootstrap your own latent: A new approach to self-supervised Learning	deepmind/deepmind-research · alibaba/EasyCV · lucidrains/byol-pytorch · +28	2020
198	SwAV (ResNet-50 x2)	77.3%	–	94M		Unsupervised Learning of Visual Features by Contrasting Cluster Assignments	open-mmlab/mmdetection · lightly-ai/lightly · facebookresearch/vissl · +15	2020
199	ReLICv2 (ResNet-50)	77.1%	–	25M		Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?	google-deepmind/relicv2	2022
200	DINO (ViT-S/16)	77.0%	–	21M		Emerging Properties in Self-Supervised Vision Transformers	facebookresearch/dino · lightly-ai/lightly · facebookresearch/vissl · +29	2021