When you gather a dataset, be sure to consider the types of Outputs you would like the final Model to produce. It is better to have a focused result in mind rather than be too broad with what you want the model to achieve. A classic model we've always loved at Scenario is called Bubbleverse. Here is the dataset:
As you can see the images are clean, clear, and focused on the style we want to achieve. For more help and guidance you can go here: Curating a Training Dataset
Go to the "Models" tab and either select the Blue Plus on left panel or click the Blue New Model button, then select Start Training:
You'll arrive at the training page and from here you can name your model and upload your images. You'll notice that each image is automatically given a description. This is called a caption. Be sure to check the captions for any errors but use caution when editing them.
Hand writing captions is an advanced practice and we recommend leaving the default captions until you've done enough research and feel confident. For more information on how to do this, visit: Guide to Manually Captioning your Dataset
We recommend for new users to start with datasets of 8 - 15 images.
We have developed the following settings that work well, but please experiment and try new things! There are many ways to create a successful training and we are always discovering new ones. We would love to see your results and experiments (the good and the bad) and one of the most enriching experiences on Scenario is the amount of space there is to discover and create!
Now you're ready to go and test your new Model! For a quick guide on this go here: Testing your Model