Classes and methods to help with the creation of geospatial training datasets and deep-learning models.
Project description
How to install geode-ml
The geode-ml package depends on GDAL and Tensorflow for most of its functionality. It is easiest to install GDAL using the conda package manager:
conda create -n "geode_env" python>=3.7
conda activate geode_env
conda install gdal
However, installing Tensorflow with Conda is trickier; we recommend following official documentation for installing the cuDNN and CUDA Toolkit libraries with the conda package manager (if you have a compatible GPU), and then doing
pip install tensorflow-gpu
on a Windows machine, or
pip install tensorflow
on a Linux machine. After activating an environment which has both GDAL and Tensorflow, use pip to install geode-ml:
pip install geode-ml
The geode.datasets module
The datasets module currently contains the classes:
- SemanticSegmentation
- creates and processes pairs of imagery and label rasters for scenes
The geode.losses module
The losses module contains custom loss functions for model training; these may be removed in the future when implementations are added to Keras.
The geode.metrics module
The metrics module contains useful metrics for testing model performance.
The geode.models module
The models module contains the classes:
- SegmentationModel
- A class which contains methods to compute metrics on test datasets,
- Unet
- a subclass of the SegmentationModel class which instantiates a Unet architecture.
- VGG19UNet
- a subclass of the SegmentationModel class which instantiates a Unet architecture, but which mirrors the VGG19 architecture for its downsampling and upsampling paths.
The geode.utilities module
The utilities module currently contains functions to process, single examples of geospatial data. The datasets module imports these functions to apply to batches of data; however, this module exists so that methods can be used by themselves, without instantiating a class object from another module.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for geode_ml-2.3.10-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | f6c5702777b9a5006060c0fe5c152473d98f89e195d780e090b4f0ddf6c7b0eb |
|
MD5 | 7a4d5d16f6e5c9fd6e298344a2eda05c |
|
BLAKE2b-256 | 648360aa5db47cf8ba9a793b9d2ca799505f5a2b4833b58e24602160c8a5e3cd |