Real-time image generation using progressive pose transfer, GAN and Spark Streaming
The model proposed in this paper gives a latest generative adversarial network for pose transfer i.e shifting the given person image to target pose in real time, target pose can be inputted through webcam/CCTV/image. In this study, we present a real-time approach to detecting people posing in an picture in 2D. The architecture consist of a chain of “Pose-Attentional Transfer Blocks” and “Open pose” Component to convert the image to pose and by using Kafka and spark to process the final poses in real time increases the performance and throughput significantly. The generated output can be fed as input to humanoid robot or can used as threat detection. The model knowledgeable and examined by the usage of Market 1501. In addition, the proposed framework will produce training images to mitigate insufficiency of evidence, re-identification of individuals.