Artists' Views on Robotics Involvement in Painting Productions

We then prompt-engineered GPT-4 (ChatGPT 4.0) to generate descriptive text labels for each image, creating 300 image-text labels for training. The CoFRIDA framework provided stroke-level painting data by simulating the robot's painting process, generating both partial-progress and completed canvases [1], [11]. While CoFRIDA is originally designed to produce paintings that accurately depict their subject matter with coherent stroke patterns and layering, our work adapts this approach specifically for abstract art by fine-tuning the InstructPix2Pix Module on a curated dataset of abstract works from Kaggle paired with GPT -4-generated captions. This specialization enables the model to generate abstract-style transformations suited for robotic execution in iterative and collaborative painting scenarios. I. Artistic Framework The robot was programmed to draw circle-inspired shapes in the first session, square-inspired in the second, and triangle-inspired in the third. This setting was inspired by Bruno Munari's [17] exploration of geometric forms (see Figure 6). In the 1960s, Italian designer Bruno Munari published visual case studies on Circle, Square, and later Triangle, associating each with specific qualities: the circle with the Divine, the square with safety, and the triangle as a key connective form.