carlosuc3m/macos-test

macos-test's Issues

Introducing the Java Deep Learning Library - JDLL

The Icy team with the help of the deepImageJ team and the Bioimage.io developers is thrilled to announce the release of JDLL, the Java Deep Learning Library. JDLL enables users to make inference with Pytorch, Tensorflow or Onnx Deep Learning models in a fast and simple manner.

JDLL is a library oriented to Java developers that want to incorporate Deep Learning models to their software. More information about it can be found at: https://github.com/bioimage-io/JDLL

With JDLL any model from any supported Deep Learning framework can be run on the same session. In addition to this, all the models and tensors are created in the same manner (using ImgLib2 in the backend) avoiding framework specific syntax.

Key features

• DL framework compatibility: JDLL currently supports Tensorflow 1, Tensorflow 2, Pytorch 1, Pytorch 2 and Onnx. Different models of different frameworks can be called in the same session. In addition, JDLL architecture will enable any new DL framework to be incorporated.

• Common model and tensor creation: each of the different DL frameworks uses different objects to load models and tensors. JDLL removes this complexity for the developer, all the models and tensors are created in the same way. The library then manages the data accordingly for each of the frameworks

• Bioimage.io compatibility: JDLL was initially developed to load models from the Bioimage.io repository. This connection has been deeply mantained and JDLL provides many methods to download and run easily Bioimage.io models. It even supports the retrieval of information of models existing in the online repository.

The library also comes with several methods to manage the installation of DL frameworks. For more information about the possibillities of JDLL, please visit the wiki: https://github.com/bioimage-io/JDLL/wiki

In order to incorporate JDLL to your project, just add the following Maven dependency and it will be ready to be used:

<dependency>
  <groupId>io.bioimage</groupId>
  <artifactId>dl-modelrunner</artifactId>
  <version>0.3.7</version>
</dependency>

With JDLL loading a Bioimage.io model is as simple as:

String bmzModelFolder = "/path/to/bioimageio/model/folder";
Model model = Model.createBioimageioModel(bmzModelFolder);
models.load();

Loading models that are not from the Bioiamge.io is only a little bit more complex. It requires defining the DL framework.

Installing a Deep Learning engine is also fairly easy:

String framework = "pytorch";
String version = "1.13.1";
boolean supportsCPU = true;
boolean supportsGPU = true;
EngineManagement.installEngineWithArgsIn(framework, vesion, supportsCPU, supportsGPU);

And downloding a model:

BioimageioRepo br = BioimageioRepo.connect();
String modelName = "Neuron Segmentation in EM (Membrane Prediction)";
br.downloadByName(modelName);

For more complete examples: https://github.com/bioimage-io/JDLL#examples

Could you please change your name?

HEllo @JDll I am writting you to ask you if you please could change your github name?
I would really appreciate it.
Regards,
Carlos