Rotational multipyramid network with bounding‐box transformation for object detection

Abstract

The study proposes a rotational multipyramid network (RoMP Net) with bounding‐box transformation for object detection. The RoMP Net is a single‐stage object detection neural network featuring three characteristics. First, the network uses a rotational bounding box to minimize the effect of background images when extracting features of objects. Bounding‐box transformation was proposed to compensate for the limitation of the rotational bounding boxes, which have relatively low prediction accuracy for objects with a high aspect ratio. Second, the RoMP Net introduces a multi‐scale and multi‐level feature pyramid network to extract distinct and semantic features efficiently. This network architecture ensures high prediction accuracy and robustness regardless of the size and complexity of objects. Third, hyperparameters in the bounding boxes are automatically determined through an unsupervised clustering method. This optimization method is also critical in improving accuracy. The performance of the proposed network and preprocessing methods are validated through image‐sets comprising critical components in power transmission facilities, which have a variety of sizes and aspect ratios. This case study demonstrates the effectiveness and robustness of the three key characteristics in the RoMP Net. Furthermore, the RoMP Net outperforms other state‐of‐the‐art deep neural networks in prediction accuracy and robustness for object detection. Specifically, the mean average precision of the RoMP Net in the validation image‐sets shows that it has the highest prediction accuracy, whereas its values in the test image‐sets confirm the network's robustness. The fast yet accurate RoMP Net will expand the range of object detection through deep neural networks.

Task‐aware swapping for efficient DNN inference on DRAM‐constrained edge systems

“…Object detection and classification are widely adopted in computer vision tasks including image segmentation, 13 event detection, 14 object tracking, 15 and so on. To deploy the deep learning inference models, fitting the resource constraints is one of the most challenging issues for edge‐end embedded devices 16,17 .…”

Active forgetting via influence estimation for neural networks

“…as it cannot be stored in memory and much less invert it for models with large parameters. For a model with parameters 2 space to calculate the Hessian. For models with a small number of parameters, the matrix can be pre-computed and stored explicitly, such that continuous requests for unlearning only involve a simple matrix-vector multiplication.…”

“…The application of machine learning as a service has shown superior performance in various applications over diverse scenarios 1,2 . User data is employed in data analysis for various tasks such as disease diagnosis, personalized recommendations, 3 and credit scoring.…”

Document images forgery localization using a two‐stream network

“…Considering the excellent performance of deep learning technologies, 30,31 the deep learning is utilized to capture forgery traces. Relying on the deep analysis of forgery methods, different traces have been utilized for forgery localization, such as noise level, 3,5,32,33 unnatural boundaries, 34 contrast/brightness inconsistencies, [35][36][37] pixels periodic correlation caused by interpolation, 1,3,19 and so on.…”