GANCCRobot: Generative adversarial nets based chinese calligraphy robot

Abstract

Robotic calligraphy, as a typical application of robot movement planning, is of great significance for the inheritance and education of calligraphy culture. The existing implementations of such robots often suffer from its limited ability for font generation and evaluation, leading to poor writing style diversity and writing quality. This paper proposes a calligraphic robotic framework based on the generative adversarial nets (GAN) to address such limitation. The robot implemented using such framework is able to learn to write fundamental Chinese character strokes with rich diversities and good quality that is close to the human level, without the requirement of specifically designed evaluation functions thanks to the employment of the revised GAN. In particular, the type information of the stroke is introduced as condition information, and the latent codes are applied to maximize the style quality of the generated strokes. Experimental results demonstrate that the proposed model enables a calligraphic robot to successfully write fundamental Chinese strokes based on a given type and style, with overall good quality. Although the proposed model was evaluated in this report using calligraphy writing, the underpinning research is readily applicable to many other applications, such as robotic graffiti and character style conversion.

Solving Robotic Trajectory Sequential Writing Problem via Learning Character’s Structural and Sequential Information

“…Based on the cosine similarity, the fitness function expressing the quality of the writing result is defined as: (14) where f denotes the fitness value; ψ(•) is the cosine similarity function; P and P ′ are vectors transformed from a numeral sample and the corresponding robotic writing result, respectively; Q and Q ′ are vectors representing the ground truth writing sequence and the trajectory sequence generated by the proposed GRU-based generator; M and M ′ the vectors obtained by centralizing Q and Q ′ , respectively; and α, β, and γ are weights to represent the significance of the three types of losses. In the experiment, α, β and γ are set to 0.2, 0.5, and 0.3, respectively.…”

“…i ) of the robot by inverse kinematics calculation, using the approach as specified in the work of [14].…”

CCD-BSMG: Composite-Curve-Dilation-Based Brush Stroke Model Generator for Robotic Chinese Calligraphy

“…Existing intelligent robot writing mainly includes stroke models based on GAN, RNN and LSTM [8], [9], [20]- [23], Auto-Encoder [24] and deep reinforcement learning [6], [7], [16]- [18].…”

Intelligent fault diagnosis of double-aligned domain adaptation based on multi-structure fusion and multi-attention mechanism

“…Qian et al [16] combined MMD with CORAL to enhance domain adaptive ability, constructed a new I-Softmax loss, and then proposed a deep discriminative TL network to realize fault diagnosis. Inspired by the generative adversarial network [17], Ganin et al [18] proposed a domain discriminator that distinguishes between source and target domain and then achieves domain confusion through adversarial training between feature extractor and domain discriminator. Wang et al [19] proposed a deep adversarial DA network, which uses domain adversarial training based on Wasserstein distance, and also combines with instance supervision to mine discriminative features with better intra-class compactness and inter-class dispersion, and better learn domain-invariant features.…”