Project Creation

Select your project's first model type

First, you create your project and define the use case. The current uses cases that are supported are

• Zero-Shot Classifier

• Few-Shot Classifier

• Classic Classifier

The main difference between those use cases is

a) the amount of data you need to upload

b) the time it will take to train your model

c) the process steps that are relevant

• For the Zero-Shot Classifier, you don't need to upload any image data. It is trained by using text descriptions of the labels that you want to recognize. The training time will take 90 seconds on average. You can completely skip Step 2 and 3.

• For the Few-Shot Classifier, you only need 2 to max 10 images per label and the training time can be as little as 3 minutes, but can go to 10-15 minutes in specific cases. In most cases, you should be able to completely skip Step 3, since there is so little data, that no tagging and curation is needed.

• For the Classic Classifier, you should expect at least 40 images per label and a training time of at least 15-20 minutes on average. If you don't need to tag your data and your data is already curated, you can skip Step 3.

Recommendation:

In general, we would recommend to start with a Zero Shot Classifier, to get quickly to a first version of your SDK and then iterate from there. If needed, you can later switch to a Few-Shot Classifier or Classic Classifier within the same project if the quality of the Zero-Shot Classifier is not sufficient. If you already have a very specific use case in mind, with an existing (medium to large) dataset that you can use as train and test data, starting with the Classific Classifier directly is probably more convenient.

If at any point you need help from the Passio team, please reach out to us at [e-mail]