Shared Visual Representations of Drawing for Communication: How do different biases affect human interpretability and intent?

Abstract

We present an investigation into how representational losses can affect the drawings produced by artificial agents playing a communication game. Building upon recent advances, we show that a combination of powerful pretrained encoder networks, with appropriate inductive biases, can lead to agents that draw recognisable sketches, whilst still communicating well. Further, we start to develop an approach to help automatically analyse the semantic content being conveyed by a sketch and demonstrate that current approaches to inducing perceptual biases lead to a notion of objectness being a key feature despite the agent training being self-supervised.

References in corpus (6)

• Learning Transferable Visual Models From Natural Language Supervision
• The Unreasonable Effectiveness of Deep Features as a Perceptual Metric
• Multi-Agent Cooperation and the Emergence of (Natural) Language
• Differentiable Drawing and Sketching
• Learning to Draw: Emergent Communication through Sketching
• From Language Games to Drawing Games