Are you interested in automatic image analysis, curious about deep learning applications, do you love geospatial data? You may appreciate the last release of our deep learning framework dedicated to image semantic segmentation!
Here comes the
0.5 release of our R&D project!
In the last version of this open source project, some appealing new features have been proposed.
Some new state-of-the-art deep neural network architectures have been implemented to enrich the existing collection, and design more sophisticated models. If the model that you are expecting is not implemented, you can anyhow take advantage of the Keras API to rebuilt it.
Furthermore a bunch of Jupyter notebooks has been written to make the framework usage easier, and clarify deep learning pipelines, from dataset description to model training and inference.
And last but not least, a light Flask Web application has been developed to showcase some deep learning predictions. Oslandia hosts this Web app at data.oslandia.io/deeposlandia. Feel free to visit it and make us some feedback!
How to use the code?
The project is still as easy to use. Clone the github repo and install it on your system with the following commands:
$ git clone https://github.com/Oslandia/deeposlandia $ cd deeposlandia $ virtualenv -p /usr/bin/python3 venv $ source venv/bin/activate (venv)$ pip install -r requirements-dev.txt
By taking the
AerialImage dataset, we have:
python deeposlandia/datagen.py -D aerial -s 250
Then a deep learning model may be trained:
python deeposlandia/train.py -M semantic_segmentation -D aerial -s 250 -e 5
This produces a trained model saved as a .h5 file on the file system. This backup may be recovered to train the model on more periods and/or to predict some image labels as follows:
python deeposlandia/inference.py -M semantic_segmentation -D aerial -i picture.png
These commands are highly configurable, do not hesitate to take a look at the README file of the project on Github.
We will be pleased to interact with you about this project, so feel free to contact us by email (email@example.com) or through Github issues! The next killing feature is maybe the one you will submit to us!