A Survey of Image Gradient Inversion Against Federated Learning

In order to preserve data privacy while fully utilizing data from different owners, federated learning is believed to be a promising approach in recent years. However, aiming at federated learning in the image domain, gradient inversion techniques can reconstruct the input images on pixel-level only by leaked gradients, without accessing the raw data, which makes federated learning vulnerable to the attacks. In this paper, we review the latest advances of image gradient inversion techniques and evaluate the impact of them to federated learning from the attack perspective. We use eight models and four datasets to evaluate the current gradient inversion techniques, comparing the attack performance as well as the time consumption. Furthermore, we shed light on some important and interesting directions of gradient inversion against federated learning.<br>

A Survey of Image Gradient Inversion Against Federated Learning

In order to preserve data privacy while fully utilizing data from different owners, federated learning is believed to be a promising approach in recent years. However, aiming at federated learning in the image domain, gradient inversion techniques can reconstruct the input images on pixel-level only by leaked gradients, without accessing the raw data, which makes federated learning vulnerable to the attacks. In this paper, we review the latest advances of image gradient inversion techniques and evaluate the impact of them to federated learning from the attack perspective. We use eight models and four datasets to evaluate the current gradient inversion techniques, comparing the attack performance as well as the time consumption. Furthermore, we shed light on some important and interesting directions of gradient inversion against federated learning.<br>

Automatic Recognition Method of Letter Images in English Self-Learning Based on Partial Differential Equation Method

According to the current situation of knowledge popularization, students simply rely on the knowledge learned in the classroom that is far from adapting to the development of modern society; so, every student needs to have the consciousness and ability of independent learning. The research of the English self-help learning system based on partial differential equation method comes into being with information network technology as the foundation for survival and development. The existing partial differential equation recognition models based on average curvature motion are all edge-based and need to use the external force defined by the image gradient to attract the zero level set (evolution curve) to move to the target edge and finally stay on the target edge. Therefore, it is difficult to obtain ideal results when extracting fuzzy or discrete boundaries (perceptual boundaries), and it is very sensitive to the selection of initial contour and noise. To solve this problem, this paper proposes a new recognition model of partial differential equations based on mean curvature motion. This overcomes some defects of existing edge models because it is region-based and does not require image gradient as a condition to stop evolution. The proposed model can avoid manual initial curve selection and allow stopping conditions to be set in the algorithm. In addition, in the numerical solution of partial differential equations, the existing model uses upwind difference scheme, and the semi-implicit additive operator separation method is adopted in this paper. Some other layers are added, and some hyperparameters are adjusted when the convolutional neural networks of inception PDEs are constructed by stacking the structure of inception PDEs. In the contrast experiment with the prototype, the software and hardware environment are the same, and the input is exactly the same. For the handwritten English alphabet data set, the variant structure can obtain more than 90% of the training accuracy and verification accuracy, which is better than the experimental accuracy of the prototype. In addition, because the inception PDE structure contains fewer parameters than the prototype, it is more computationally efficient and takes less training time per batch than the prototype.

Image Preprocessing for Artistic Robotic Painting

Artistic robotic painting implies creating a picture on canvas according to a brushstroke map preliminarily computed from a source image. To make the painting look closer to the human artwork, the source image should be preprocessed to render the effects usually created by artists. In this paper, we consider three preprocessing effects: aerial perspective, gamut compression and brushstroke coherence. We propose an algorithm for aerial perspective amplification based on principles of light scattering using a depth map, an algorithm for gamut compression using nonlinear hue transformation and an algorithm for image gradient filtering for obtaining a well-coherent brushstroke map with a reduced number of brushstrokes, required for practical robotic painting. The described algorithms allow interactive image correction and make the final rendering look closer to a manually painted artwork. To illustrate our proposals, we render several test images on a computer and paint a monochromatic image on canvas with a painting robot.