Sketch-RNN

Draw a basic shape representing the icon.

Select a relevant model (e.g., 'arrow', 'button').

Let Sketch-RNN generate multiple variations.

Select the most promising variation and refine it.

Draw a random shape or line.

Select a diverse model (e.g., 'cat', 'flower').

Observe the generated completions.

Use the generated image as a starting point for a new artwork.

Students experiment with drawing and model selection.

Observe how the neural network learns and generates sketches.

Discuss the underlying principles of recurrent neural networks and variational autoencoders.

Analyze the strengths and limitations of the model.

Start drawing a basic character outline.

Select a model related to the character type (e.g., 'animal', 'person').

Let Sketch-RNN generate variations of the character's features.

Combine the generated features to create a unique character design.

Draw a symbolic representation of the concept.

Select a model that aligns with the concept's theme (e.g., 'cloud' for 'dream', 'gear' for 'technology').

Observe the generated completions.

Interpret the generated images as visual metaphors for the concept.

Sketch a basic shape for the logo.

Select a fitting model (e.g., 'flower', 'tree').

Review the generated completions for inspiration.

Refine chosen design manually.

Repeat until suitable draft is created.