Image cropping is a crucial means of enhancing the aesthetic quality of images. However, it relies on photographic knowledge and experience. For novice users, it is challenging to obtain satisfactory results through cropping. To meet the image cropping needs of novice users and improve efficiency, this article designs and develops an intelligent image cropping application, AesthetiCropper, and proposes two cropping methods. One is based on visual balance. The other is based on deep learning. This system features an intuitive user interface and automatically crops images to high aesthetic standards. It also offers customizable parameters and an editing module to meet users’ personalized requirements. User studies show that participants are more satisfied with the system’s cropping results compared to manual ones. The System Usability Scale (SUS) score indicates that its usability reaches a good level, making it worthy of promotion.

Image cropping is an important way to improve the composition of images, and it depends on aesthetic literacy and photography skills, which are the outcomes of years of training and practice. For novice users who lack photography knowledge, it’s difficult to get well-composed cropping results, and there is no application-level solution to help them. We develop an image cropping software MAGICROP, which is designed for novice users to make artful portrait images automatically. With an intuitive user interface, the software is enabled with both deep learning and composition rules for cropping operations. From the two functional models provided by the system, users can get inspiration of cropping and set customized parameters to generate the required cropping results. User studies verify the effectiveness of MAGICROP in creating quality images. The SUS score suggests acceptable usability and a promising space for continuous improvement.

A robotic puppet is developed for studying motion generation and control of highly articulated biomimic mechatronic systems with anatomical motion data of human in real time. The system is controlled by a pair of data gloves tracking human fingers' actions. With the primitives designed in a multilayered motion synthesis structure, the puppet can realize some complex human-like actions. Continuous full body movements are produced on the robotic puppet by combining and sequencing the actions on different body parts using temporal and spatial information provided by the data gloves. Human is involved in the interactive design of the coordination and timing of the body movements of the robotic puppet in a natural and intuitive manner. The methods of motion generation exhibited on the robotic puppet may be applied to the interactive media, entertainment and biomedical engineering.

This paper describes the generation of expressive gestures on a robotic puppet controlled using behavior-based method. Recent findings on biological motion mechanisms provide an inspiring model for controlling high dimensional robot with motor primitives. The modular features of primitives make them the building blocks for constructing complex behaviors. The inborn similarity between gesture and speech enable us to study behavior with those methods for language. We designed a motor controller using a small set of primitives. The effectiveness of this method is validated on the robotic puppet with different expressive gestures generated by superimposing and sequencing the primitives.