Introduction of Image Recognition Packages Project that based on Sentis

About

Hi Unity Sentis users,
We have released an image recognition packages that implemented based on Unity Sentis.
You can easily to use class classification, object detection, and other with this package.
It is OSS licensed under the MIT License and you can be used for commercial.
This packages will help you for implement image recognition in your app. Thanks,




Setup

The library consists of one base package in inference features are implemented, and several tasks packages in which image recognition are implemented.
Each task package depends on the base package and should be imported as a set with the base package.
You can easily setup these packages from package manager.

e.g. Setup for Classification

  • Package Manager
    1. [Window]>[Package Manager]
    2. [+]>[Add package from git URL…]
    3. Add the following URLs
      • https://github.com/HoloLabInc/HoloLabDnnPackages.git?path=packages/jp.co.hololab.dnn.base
      • https://github.com/HoloLabInc/HoloLabDnnPackages.git?path=packages/jp.co.hololab.dnn.classification

Scripts

You can implement image recognition with just writing a few lines of script.
The task packages are includes a simple sample app to learn how to use.
Please see sample for details.

  1. Reference Namespaces
 using Unity.Sentis;
 using HoloLab.DNN.Classification;
  1. Create Instance
[SerializeField] private ModelAsset weights; // attach in inspector
var model = new ClassificationModel(weights);
  1. (Option) Set Parameters
model.SetApplySoftmax(true);
  1. Run Inference
(var class_id, var score) = model.Classify(texture);
3 Likes

That’s great! I think people will find this very useful! :metal: Nice work @SugiuraTsukasa

BTW, we have found that a good place to bring your implementations to a wider audience is on Hugging Face. That is where most people look for models. Just upload your models (either ONNX or Sentis files) and choose the “Unity Sentis” library tag. Then you can link to your github packages in the readme file.

BTW, how did you find using Unity Sentis? Did you have any problems or was it all smooth sailing? :sailboat:

1 Like