# Clip-fine-tuned-satellite Open-source Model - High-precision Identification of Satellite Images with an Accuracy of 96.9%

Clip Fine Tuned Satellite

Developed by NemesisAlm

A fine-tuned version of the CLIP model on the UC_Merced satellite image dataset, achieving 96.9% accuracy

Open Source License:MIT #Satellite Image Classification #High-Precision CLIP Fine-Tuning #Remote Sensing Image Recognition

Downloads 30

Release Time : 8/21/2024

Model Overview

This model is used for satellite image classification, supporting the recognition of 21 types of land cover, with significant performance improvements over the original CLIP model

Model Features

High-Precision Classification

Achieves 96.9% accuracy on the UC_Merced test set, a 38 percentage point improvement over the original CLIP model

Efficient Fine-Tuning

Only 30% of parameters were trained, achieving performance leaps in just 2 epochs

Multi-Scenario Coverage

Supports classification of 21 satellite image scenarios, including farmland, forests, residential areas, etc.

Model Capabilities

Satellite Image Classification

Multi-Label Image Recognition

Cross-Modal Understanding (Image-Text)

Use Cases

Geographic Information Systems

Land Use Monitoring

Automatically identifies land use types in satellite images

Accurately distinguishes 21 types of land cover

Urban Planning

Analyzes the density distribution of urban areas

Can identify high/medium/low-density residential areas

Environmental Monitoring

Forest Cover Assessment

Monitors changes in forested areas

Accurately identifies forest and shrubland types

🚀 clip-fine-tuned-satellite

This model is a fine-tuned version of CLIP, which significantly improves the accuracy of satellite image classification.

🚀 Quick Start

This model is a fine-tuned version of openai/clip-vit-base-patch32 on the blanchon/UC_Merced dataset. It achieves the following results on the test set:

Accuracy: 96.9% The original CLIP model achieves 58.8% of accuracy.

✨ Features

High Accuracy: After fine-tuning, the model achieves an accuracy of 96.9% on the test set, significantly higher than the original CLIP model.
Efficient Training: Only 2 epochs of training were needed, with 30% of the parameters retrained to achieve a significant increase in accuracy.

📚 Documentation

Model description

The model is a fine-tuned version of CLIP. 30% of the parameters were retrained to achieve a significant increase in accuracy after only 2 epochs.

Intended uses & limitations

The model is to be used to classify satellite images. It was trained on the UC_Merced dataset that comprises 21 classes: agricultural, airplane, baseballdiamond, beach, buildings, chaparral, denseresidential, forest, freeway, golfcourse, harbor, intersection, mediumresidential, mobilehomepark, overpass, parkinglot, river, runway, sparseresidential, storagetanks, tenniscourt.

To see how to use it, refer to the CLIP documentation or check the app using this model: https://huggingface.co/spaces/NemesisAlm/clip-satellite-demo

Training and evaluation data

30% of the parameters trained. Evaluated against a test set of 420 images.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 3e-05
train_batch_size: 64
eval_batch_size: 64
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 2
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
1.4974	1.0	20	3.0190
1.3733	2.0	40	2.9588

Framework versions

Transformers 4.44.0
Pytorch 2.4.0+cu121
Datasets 2.20.0
Tokenizers 0.19.1

📄 License

This model is released under the MIT license.

📦 Model Information

Property	Details
Base Model	openai/clip-vit-base-patch32
Model Name	clip-fine-tuned-satellite-20240821
Tags	generated_from_trainer
Datasets	blanchon/UC_Merced
Metrics	accuracy
Library Name	transformers

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご