Creating a new dataset
All models (algorithms) in Criterion AI need a dataset to be trained on. For that reason, the first step in Criterion AI is always to create a representative dataset that follows the structure required by the model(s) you wish to train. You can learn more about the data structure by reading the instructions of the model type that you wish to base your model on. In the example below, we can see that, in order to train a model based on the Image Classification model type, the dataset must consist of images structured in folders designating the categories (classes) of the images.
With this structure in mind, we can now create a new dataset. Navigate to the Data section and click on the button Create a new dataset to open the wizard for creating a new dataset.
Start out by giving your dataset a name. It can be a good idea to give your datasets names that let you easily identify them in the future. Also, you might want to create separate datasets for training purposes and for test purposes. You can indicate the purpose of the dataset in its name.
Once you have chosen a suitable name, click on the Next button and use the File Explorer to upload data to your dataset. See the article on uploading data to learn more about how that process works.
After having selected the data you wish to upload, click on the Start data transfer button to create your dataset and to initiate the upload process of your files.