This paper presents a technical account of the development of an the augmented architectural drawing tool Sketch2Pix: an interactive application that sup-ports architectural sketching augmented by automated image-to-image translation processes.

Here, we account not only for the "front-end" experience of this tool, but also detail our approach to offering novice designers access to the "full stack" of processes and technologies required to craft and interact with their own augmented drawing assistant.

This workflow includes: the establishment of a training data set; the training and validation of a neural network model; the deploying of this model in a graphic sketching environment; and the configuration of the sketching environment to facilitate a "conversation" with an AI partner. Our approach is novel both in the technical barriers it removes, and in the facilitation of access for creative design practitioners who may not be well-versed in the underlying technologies.

The accessibility of our approach is demonstrated by a case study in the Spring of 2020. At this time, a group of novice undergraduate students of design at UC Berkeley adopted Sketch2Pix to produce a series of image-to-image translation models, or "brushes", each trained on a designer's own data, and for a designer's own use in sketching architectural forms. By empowering creative practitioners to more easily shape the configuration of machine-augmented tools, this project seeks to serve as a harbinger of novel modalities of authorship that we might expect in this still-emerging paradigm of computer-assisted design.

Read the full paper here.