NWPU-RESISC45

NWPU-RESISC45

MH Specific

Large-scale remote sensing image scene classification benchmark with 45 scene classes extracted from Google Earth covering 100+ countries. Applicable to Maharashtra urban planning, land use mapping, and infrastructure analysis

Fine-tune a remote sensing image classifier pre-trained on NWPU-RESISC45 for Maharashtra land use classification.
HomepageHuggingFace

Quick Start

# NWPU-RESISC45 dataset
from torchvision import datasets
print('NWPU-RESISC45: 45-class remote sensing image classification')
print('31,500 images at 256x256 resolution')
Modality
Satellite imagery (RGB)
Size
31.5K images (256x256 px), 45 classes
License
Format
PNG/JPEG
Language
mr
Update Frequency
static
Organization
Northwestern Polytechnical University, China
Category
vision-ocr-multimodal

Schema

FieldTypeDescription
imageimageRemote sensing scene image (256x256 pixels)
labelstringScene class label (45 classes)

Build With This

Create a transfer learning pipeline from NWPU-RESISC45 to Maharashtra-specific satellite image classification tasks
Develop a few-shot scene classifier for rare land use types in Maharashtra using representations learned from RESISC45
Build an automated satellite image annotation tool using RESISC45-trained models as initial labelers

AI Use Cases

Remote sensing scene classificationurban planning
Last verified: 2026-03-07