Getting started with Einstein Object Detection
To get started with Einstein Vision, open the Einstein Vision Projects tab.
Create a new Einstein Vision Project and select Object Detection in the Record Type menu and do Next.
Fill in the required fields and save the record.
In the Related tab on the record, you will see three related objects:
Einstein Vision Labels
- These are the labels that will be used to assign to bounding box
- This object will contain the album on which you will be uploading images and drawing bounding box. You can create how much you want and insert how much images required.
Einstein Vision Datasets
- Once you are satisfied with the label, you may proceed to create a dataset and eventually later a model.
Einstein Vision Labels will contain all the labels that you want when labeling objects in your images.
Start by creating your labels.
Once the labels have been created, you can proceed with labeling images. This process requires you to create Boxing Tasks
Note: it is required before you can move on.
This record will contain the interface on which you can start labelling objects on images.
Create how much Boxing Tasks required and add images. Ideally, a set of 20 images per Boxing Task will make you go fast and easily dispatched among your team members.
To start labelling, click on the Einstein Label icon. It will appear when you click on an image.
Draw bounding boxes around objects that you want to detect and make sure you associate them with the appropriate labels.
Creating the Einstein Vision Dataset, will create a snapshot of all the boxing tasks with images and bounding box and convert them into training data for you to be able to train a model later on.
You should see your images being displayed on the dataset album. It will take a few minutes depending on the amount of images which were on the Boxing Tasks.
Please note that you can't edit a Einstein Vision Dataset after it has been created.
On the related tab of the Einstein Vision Dataset, you can create a new Einstein Model. By creating a Einstein Model, this will start the training of the dataset and a model will be generated. This will be used to deliver the predictions.
More information regarding Advanced parameters can be found here: https://metamind.readme.io/reference#train-a-dataset
Insert advanced parameters only if you know what you are doing else your model might not perform as expected or may even fail.
Model training takes time. It all depend on the number of images and labels and bounding box contain in dataset. A progress bar is available for you to see the status over time.
When the Einstein Model training has been completed successfully. You will be provided with a SharinPix album where you can add images. Upon opening the images, a prediction call will be made and in a few seconds, you should be able to see the results of the prediction on the image itself.
Contact SharinPix Support on the following email if you require assistant getting started: [email protected]