---
title: >-
  DeepSlide - Landslide Detection and Mapping Using Deep Learning Across
  Multi-Source Satellite Data and Geographic Regions
emoji: 🌍
colorFrom: blue
colorTo: green
sdk: docker
app_port: 8501
tags:
- streamlit
- pytorch
- deep-learning
- landslide-detection
pinned: true
short_description: Landslide Detection using Deep Learning
license: apache-2.0
paper:
  - https://huggingface.co/papers/2507.01123
---

# DeepSlide: Landslide Detection Models

This Space demonstrates various deep learning models for landslide detection, using models trained with PyTorch. The models are served directly from our [harshinde/DeepSlide_Models](https://huggingface.co/harshinde/DeepSlide_Models) or [Kaggle Models Repository](https://www.kaggle.com/models/harshshinde8/sims/).

## Available Models
- DeepLabV3+
- DenseNet121
- EfficientNetB0
- InceptionResNetV2
- InceptionV4
- MiT-B1
- MobileNetV2
- ResNet34
- ResNeXt50_32X4D
- SE-ResNet50
- SE-ResNeXt50_32X4D
- SegFormer
- VGG16

## How to Use
1. Select a model from the sidebar
2. Upload one or more `.h5` files containing satellite imagery
3. View the landslide detection results and predictions
4. Download the results if needed

## Model Information
All models are trained on satellite imagery data and are optimized for landslide detection. Each model has its own strengths and characteristics, which are described in the app interface when you select them.

## Technical Details
- Python 3.9
- PyTorch 1.9.0
- Streamlit 1.28.0
- Models are automatically downloaded from HuggingFace [harshinde/DeepSlide_Models](https://huggingface.co/harshinde/DeepSlide_Models).
- Dataset [harshinde/LandSlide4Sense](https://huggingface.co/datasets/harshinde/LandSlide4Sense).

## Author
- Harsh Shinde

#  *DeepSlide: Landslide Detection and Mapping Using Deep Learning Across Multi-Source Satellite Data and Geographic Regions*

Landslide4sense Dataset:

```Landslide4sense Dataset
from datasets import load_dataset

ds = load_dataset("harshinde/LandSlide4Sense")
```
Deepslide Models - [harshinde/DeepSlide_Models](https://huggingface.co/harshinde/DeepSlide_Models)

Dataset [harshinde/LandSlide4Sense](https://huggingface.co/datasets/harshinde/LandSlide4Sense)

Wandb Results - https://wandb.ai/Silvamillion/Land4Sense

Paper - https://dx.doi.org/10.2139/ssrn.5225437

## 📄 Citation

If you use **DeepSlide-L4S-Code** or reference our work in your research, please cite our paper:

> Harsh Shinde, et al. *Landslide Detection and Mapping Using Deep Learning Across Multi-Source Satellite Data and Geographic Regions*, SSRN, 2024.  
> DOI: [10.2139/ssrn.5225437](https://dx.doi.org/10.2139/ssrn.5225437)

BibTeX:
```bibtex
@article{burange2025landslide,
  title={Landslide Detection and Mapping Using Deep Learning Across Multi-Source Satellite Data and Geographic Regions},
  author={Burange, Rahul and Shinde, Harsh and Mutyalwar, Omkar},
  journal={Available at SSRN 5225437},
  year={2025}
  eprint={5225437},
  archivePrefix={SSRN},
  doi={10.2139/ssrn.5225437},
  url={https://dx.doi.org/10.2139/ssrn.5225437}
}