Scailable X Edge Impulse: Effortless model training and deployment

We are very excited to announce our integration with Edge Impulse. Edge Impulse provides an amazing platform for capturing and annotating training data and for training highly accurate AI models using advanced transfer learning tools. Edge Impulse makes model creation a breeze. By connecting your Edge Impulse account to your Scailable platform account we can now move models between Edge Impulse and Scailable.

Edge Impulse focusses on creating highly performant (both in terms of accuracy and computational speed), accurate ML and AI models. Edge Impulse makes data management and labelling easy, and through its advanced model architectures it allows for easy (re-)training of models such that accurate segmentation models can be created with only a small number of training images.

We pick it up after model training; simply connect your Edge Impulse account with your Scailable platform account to import your trained Edge Impulse model to Scailable and deploy the model, in a highly efficient AI pipeline, on Scailable supported hardware.

The connection between Edge Impulse and Scailable closes the gap to make large scale edge AI solution creation, deployment, and secure management possible without any code. Train your model using a highly intuitive UI, configure your model pipeline remotely, and deploy your model over-the-air. A process that took months off highly skilled engineering (both embedded and ML engineering) is now reduced to minutes. Scailable X Edge Impulse allows you to iterate faster, improve your AI and ML solutions, move beyond Proof-of-concepts and provide business value in the field.

Edge Impulse X Scailable, step by step.

Here we detail step-by-step how to create your first edge AI solution, start-to-end, using Edge Impulse X Scailable.

Step 1: Hardware selection

Any Edge AI solution requires appropriate hardware. The Scailable AI manager comes pre-installed on a large number of selected devices. On these devices Scailable ensure a highly performant and highly secure AI pipeline that is easily remotely configurable with all the options you need to create your edge AI solution.

Get your hands on a selected edge device, or install the Scailable AI manager on a local device for testing purposes.

Step 2: Account creation

We will use Scailable for model deployment, and Edge Impulse for model training. So, step two is to setup accounts with both Scailable and Edge Impulse.

Step 3: Model training using Edge Impulse

Once you have access to your Edge Impulse account, let the intuitive UI guide you through the process of creating a FOMO segmentation model. Or, just start from an existing project.

Step 4: Connecting your accounts

Once you have created your AI model in Edge Impulse, use the Edge Impulse project API key to connect your Edge Impulse account with your Scailable Platform account. You can now import your Edge Impulse model effortlessly.

Step 5: Effortless model deployment

Once you have imported your Edge Impulse model to Scailable, you can navigate to the AI manager installed on your selected device and configure your full AI pipeline. You simply select the input sources, sampling scheme, model, and output destination and you are good to go. The Scailable AI manager provides easy, on-device, visualizations for checking your setup.

Note: The AI manager installed on selected devices can be configured to capture new training images which can be send to Edge Impulse for model re-training. Thus, Scailable X Edge Impulse allows for quick iterations to create a highly performant edge AI solution. We will discuss more about setting up the full round-trip in a future post.

Get started.

Looking for more information? Please check out our documentation:

• AI Manager docs
• Scailable platform docs
• Edge Impulse X Scailable integration docs
• Edge Impulse docs

Looking to kickstart your Edge AI solution? We can always help, please contact our support..