HyperAI超神経

ホームプラットフォームドキュメントニュース論文チュートリアルデータセット百科事典 SOTA LLMモデル GPU ランキング学会

サイトについて

日本語

HyperAI超神経

Image Classification On Resisc45

評価指標

Top 1 Accuracy

評価結果

このベンチマークにおける各モデルのパフォーマンス結果

		Paper Title	Repository
ResNet50	96.83	In-domain representation learning for remote sensing
LWGANet L2	96.17	LWGANet: A Lightweight Group Attention Backbone for Remote Sensing Visual Tasks
DecoupleNet D2	95.87	DecoupleNet: A Lightweight Backbone Network With Efficient Feature Decoupling for Remote Sensing Visual Tasks	-
LWGANet L1	95.70	LWGANet: A Lightweight Group Attention Backbone for Remote Sensing Visual Tasks
SEER (RegNet10B)	95.61	Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision
LWGANet L0	95.49	LWGANet: A Lightweight Group Attention Backbone for Remote Sensing Visual Tasks
AGOS	94.91	All Grains, One Scheme (AGOS): Learning Multi-grain Instance Representation for Aerial Scene Classification
SwAV (ResNet50-w5)	94.73	Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision
DINO (DeiT-B/16)	93.97	Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision
LSENet	93.49	Local semantic enhanced convnet for aerial scene recognition	-
MoCo-v3 (ViT-B/16)	93.35	Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision
CLIP (ViT-B/16)	92.7	Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision
BYOL (ResNet200-w2)	92.53	-	-
DeiT-B/16	92.48	Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision
SimCLR-v2 (ResNet152-w3 + SK)	89.77	Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision
ResNet50 (ImageNet-supervised)	88.56	Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision
MIDC-Net	87.99	A multiple-instance densely-connected ConvNet for aerial scene classification	-
MoCo-v2 (ResNet50)	85.4	Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision
SAG-ViT	-	SAG-ViT: A Scale-Aware, High-Fidelity Patching Approach with Graph Attention for Vision Transformers

0 of 19 row(s) selected.