Introduction
In this tutorial series you will learn how to integrate your custom model into Supervisely by creating a simple serving app.
✅ Integration process is simple - the only thing you need is to implement a method of how your model gets prediction of an image. Supervisely SDK will handle the rest automatically.
If you are using popular machine learning frameworks, you can skip integration and start using already existing apps in Supervisely Ecosystem. Most popular neural network frameworks are already integrated into Supervisely. Users can train these models on their data and test them (inference) right in the platform in a few clicks.
We highly recommend to explore apps in Supervisely Ecosystem, here are several examples of ready-to-use frameworks for instance segmentation:
If your favorite NN framework is not in our Ecosystem yet, you can send us a feature request in Supervisely Ideas Exchange.
Benefits
Once you implement a serving application for your NN architecture, you can do a lot of things, like inference on your data for pre-labeling to speed up annotation, perform active learning, analyze and debug your model with various data science tools, combine models into pipelines and many more.
Find more use cases and video tutorials on our youtube channel.
Generally speaking, your model will be compatible with the entire ecosystem of applications in Supervisely.
Here are the examples of apps you might be interested to use with your model:
Apply NN to Images Project
app - apply NN to your images and save predictionsNN Image Labeling
app - use NN right in labeling interfaceApply Detection and Classification Models to Images Project
app - combine models into pipelinesApply NN to Videos Project
app - predict and track objects on videosInference via Session API: You can also connect to the model and get the inference in a couple of lines with the help of the
sly.nn.inference.Session
class. See our Inference API Tutorial.
We have tutorials for all CV tasks
Just pick one you need and get started:
Last updated