Adobe, Inc.

Biao Jia

Jonathan Worthen Brandt

Byungmoon Kim

Chen Fang

Radomir Mech

Ning Xu

Some embodiments involve a reinforcement learning based framework for training a natural media agent to learn a rendering policy without human supervision or labeled datasets. The reinforcement learning based framework feeds the natural media agent a training dataset to implicitly learn the rendering policy by exploring a canvas and minimizing a loss function. Once trained, the natural media agent can be applied to any reference image to generate a series (or sequence) of continuous-valued primitive graphic actions, e.g., sequence of painting strokes, that when rendered by a synthetic rendering environment on a canvas, reproduce an identical or transformed version of the reference image subject to limitations of an action space and the learned rendering policy.

Reinforcement learning-based techniques for training a natural media agent

A non-photorealistic image rendering system and related techniques are described herein that train and implement machine learning models to reproduce digital images in accordance with various painting styles and constraints. The image rendering system can include a machine learning system that utilizes actor-critic based reinforcement learning techniques to train painting agents (e.g., models that include one or more neural networks) how to transform images into various artistic styles with minimal loss between the original images and the transformed images. The image rendering system can generate constrained painting agents, which correspond to painting agents that are further trained to reproduce images in accordance with one or more constraints. The constraints may include limitations of the color, width, size, and/or position of brushstrokes within reproduced images. These constrained painting agents may provide users with robust, flexible, and customizable non-photorealistic painting systems.

Growing community of inventors

Lanham, MD, United States of America

Biao Jia