DCAP: Landslide Segmentation Checkpoints
1 min readupdated
The model implements the Detail–Context Adaptive Structured Pruning approach for precise detection and segmentation of landslides using satellite or multi-channel data. The release features two ensemble builds: Dense ResUNet and DCAP70, each utilizing probability averaging across three seeds. It accepts specialized 21-channel reflectance and change features as input. The solution is designed for researchers and specialists in remote sensing and geological hazard analysis.
Key features
- Includes Dense ResUNet and DCAP70 checkpoints for precise analysis
- Utilizes a three-member ensemble with predicted probability averaging
- Processes complex 21-channel reflectance and change input data
- Distributed under the permissive MIT license with full reproduction instructions
What you can do
- Evaluate model performance using the official project CLI tool
- Integrate the checkpoints into remote sensing data processing pipelines
- Benchmark the Dense ResUNet and DCAP70 ensembles on custom datasets
- Reproduce the results following the official repository guide
Example
hf download ghimiredhikura/DCAP --include "*.pt" --include MODEL_NOTICES.txt --local-dir downloads/models
python -m dcap models install --source downloads/modelsRead more
DetailsArchitecture: Dense ResUNet and DCAP70
Input data: 21-channel reflectance and change features
Ensembles: Probability averaging of three seed models
License: MIT










Comments