ML Research Wiki / Benchmarks / Image Classification / RESISC45

RESISC45

Image Classification Benchmark

Performance Over Time

📊 Showing 16 results | 📏 Metric: Top 1 Accuracy

Top Performing Models

Rank	Model	Paper	Top 1 Accuracy	Date	Code
1	ResNet50 📚	In-domain representation learning for remote sensing	96.83	2019-11-15	📦 google-research/google-research
2	LWGANet L2	LWGANet: A Lightweight Group Attention Backbone for Remote Sensing Visual Tasks	96.17	2025-01-17	📦 lwcver/lwganet
3	LWGANet L1	LWGANet: A Lightweight Group Attention Backbone for Remote Sensing Visual Tasks	95.70	2025-01-17	📦 lwcver/lwganet
4	SEER (RegNet10B) 📚	Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision	95.61	2022-02-16	📦 facebookresearch/vissl
5	LWGANet L0	LWGANet: A Lightweight Group Attention Backbone for Remote Sensing Visual Tasks	95.49	2025-01-17	📦 lwcver/lwganet
6	SAG-ViT	SAG-ViT: A Scale-Aware, High-Fidelity Patching Approach with Graph Attention for Vision Transformers	95.49	2024-11-14	📦 shravan-18/SAG-ViT
7	AGOS	All Grains, One Scheme (AGOS): Learning Multi-grain Instance Representation for Aerial Scene Classification	94.91	2022-05-06	📦 biqiwhu/agos
8	SwAV (ResNet50-w5) 📚	Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision	94.73	2022-02-16	📦 facebookresearch/vissl
9	DINO (DeiT-B/16) 📚	Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision	93.97	2022-02-16	📦 facebookresearch/vissl
10	MoCo-v3 (ViT-B/16) 📚	Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision	93.35	2022-02-16	📦 facebookresearch/vissl

All Papers (16)

In-domain representation learning for remote sensing

2019

ResNet50

google-research/google-research

LWGANet: A Lightweight Group Attention Backbone for Remote Sensing Visual Tasks

2025

LWGANet L2

lwcver/lwganet

LWGANet: A Lightweight Group Attention Backbone for Remote Sensing Visual Tasks

2025

LWGANet L1

lwcver/lwganet

Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision

2022

SEER (RegNet10B)

facebookresearch/vissl

LWGANet: A Lightweight Group Attention Backbone for Remote Sensing Visual Tasks

2025

LWGANet L0

lwcver/lwganet

SAG-ViT: A Scale-Aware, High-Fidelity Patching Approach with Graph Attention for Vision Transformers

2024

SAG-ViT

shravan-18/SAG-ViT

All Grains, One Scheme (AGOS): Learning Multi-grain Instance Representation for Aerial Scene Classification

2022

AGOS

biqiwhu/agos

Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision

2022

SwAV (ResNet50-w5)

facebookresearch/vissl

Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision

2022

DINO (DeiT-B/16)

facebookresearch/vissl

Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision

2022

MoCo-v3 (ViT-B/16)

facebookresearch/vissl

Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision

2022

CLIP (ViT-B/16)

facebookresearch/vissl

Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision

2022

DeiT-B/16

facebookresearch/vissl

Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision

2022

SimCLR-v2 (ResNet152-w3 + SK)

facebookresearch/vissl

Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision

2022

ResNet50 (ImageNet-supervised)

facebookresearch/vissl

Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision

2022

MoCo-v2 (ResNet50)

facebookresearch/vissl

SkySense: A Multi-Modal Remote Sensing Foundation Model Towards Universal Interpretation for Earth Observation Imagery

2023

SkySense-O

jack-bo1220/awesome-remote-sensing-foundation-models

RESISC45

Performance Over Time

Edit Benchmark Results

Edit Result

Top Performing Models

All Papers (16)

In-domain representation learning for remote sensing

LWGANet: A Lightweight Group Attention Backbone for Remote Sensing Visual Tasks

LWGANet: A Lightweight Group Attention Backbone for Remote Sensing Visual Tasks

Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision

LWGANet: A Lightweight Group Attention Backbone for Remote Sensing Visual Tasks

SAG-ViT: A Scale-Aware, High-Fidelity Patching Approach with Graph Attention for Vision Transformers

All Grains, One Scheme (AGOS): Learning Multi-grain Instance Representation for Aerial Scene Classification

Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision

Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision

Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision

Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision

Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision

Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision

Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision

Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision

SkySense: A Multi-Modal Remote Sensing Foundation Model Towards Universal Interpretation for Earth Observation Imagery

Model	Paper	Top 1 Accuracy	Date
ResNet50	In-domain representation learning for remote sens…	96.83	2019-11-15
LWGANet L2	LWGANet: A Lightweight Group Attention Backbone f…	96.17	2025-01-17
LWGANet L1	LWGANet: A Lightweight Group Attention Backbone f…	95.70	2025-01-17
SEER (RegNet10B)	Vision Models Are More Robust And Fair When Pretr…	95.61	2022-02-16
LWGANet L0	LWGANet: A Lightweight Group Attention Backbone f…	95.49	2025-01-17
SAG-ViT	SAG-ViT: A Scale-Aware, High-Fidelity Patching Ap…	95.49	2024-11-14
AGOS	All Grains, One Scheme (AGOS): Learning Multi-gra…	94.91	2022-05-06
SwAV (ResNet50-w5)	Vision Models Are More Robust And Fair When Pretr…	94.73	2022-02-16
DINO (DeiT-B/16)	Vision Models Are More Robust And Fair When Pretr…	93.97	2022-02-16
MoCo-v3 (ViT-B/16)	Vision Models Are More Robust And Fair When Pretr…	93.35	2022-02-16
CLIP (ViT-B/16)	Vision Models Are More Robust And Fair When Pretr…	92.70	2022-02-16
DeiT-B/16	Vision Models Are More Robust And Fair When Pretr…	92.48	2022-02-16
SimCLR-v2 (ResNet152-w3 + SK)	Vision Models Are More Robust And Fair When Pretr…	89.77	2022-02-16
ResNet50 (ImageNet-supervised)	Vision Models Are More Robust And Fair When Pretr…	88.56	2022-02-16
MoCo-v2 (ResNet50)	Vision Models Are More Robust And Fair When Pretr…	85.40	2022-02-16
SkySense-O	SkySense: A Multi-Modal Remote Sensing Foundation…	83.28	2023-12-15