Analysis of two streaming frameworks: Apache Storm and IBM InfoSphere Streams was performed in solving a multiple object detection task. The analysis focused on two parameters: throughput and 95th percentile of image processing delay. Faces were chosen as the objects to be detected. Profiling was held under CentOS operating systems running on a five node cluster. Face detection was performed using an OpenCV cascade classifier. First architectures and the experiment description were covered. Final suggestions on the applicability of the two systems were made in the concluding section of the article. Apache Storm demonstrated a scalability advantage over IBM InfoSphere Streams in the experiment conducted. It was confirmed that the system based on Apache Storm was able to operate on FullHD video in real-time, achieving throughput of 24 images per second on the hardware used.

Comparison of two streaming systems: Apache Storm and IBM InfoSphere Streams was performed in solving image sliding window filtering task. Analysis was focused on two parameters: throughput and memory consumption. Profiling was held under CentOS operating systems running on two virtual machines for each system. First, the most cheap in developing cost realizations were compared. After analysis optimization of memory consumption was performed. Final suggestions on applicability of two systems were made at the end of the article.

This paper examines the effectiveness of neural network algorithms for hydraulic system fault detection and a novel neural network architecture is suggested. The proposed gated convolutional autoencoder was trained on a simulated training set augmented with just 0.2% data from the real test bench, dramatically reducing the time needed to spend with the actual hardware to build a high-quality fault detection model. Our fault detection model was validated on a test bench and showed accuracy of more than 99% of correctly recognized hydraulic system states with a 10-s sampling window. This model can be also leveraged to examine the decision boundaries of the classifier in the two-dimensional embedding space.