Improved Mask R-CNN Multi-Target Detection and Segmentation for Autonomous Driving in Complex Scenes

Review Computer Science, Software Engineering

Attention mechanisms in computer vision: A survey

Meng-Hao Guo et al.

Summary: Attention mechanisms, inspired by the human visual system, have been successfully applied in various computer vision tasks. This survey provides a comprehensive review of different types of attention mechanisms and suggests future research directions.

COMPUTATIONAL VISUAL MEDIA (2022)

添加到收藏夹

Article Automation & Control Systems

Enhancing Geometric Factors in Model Learning and Inference for Object Detection and Instance Segmentation

Zhaohui Zheng et al.

Summary: The proposed CIoU loss and Cluster-NMS approach, which incorporates geometric factors, significantly improve average precision and average recall in object detection and instance segmentation, with notable gains without sacrificing inference efficiency.

IEEE TRANSACTIONS ON CYBERNETICS (2022)

添加到收藏夹

Proceedings Paper Computer Science, Artificial Intelligence

E2EC An End o-End Contour-based Method for High-Quality High-Speed Instance Segmentation

Tao Zhang et al.

Summary: This paper presents a novel contour-based method called E2EC for high-quality instance segmentation. By using a learnable contour initialization and a dynamic matching loss function, E2EC improves the accuracy and boundary detail quality of the segmentation. Experimental results demonstrate that E2EC achieves state-of-the-art performance on multiple datasets and is efficient for real-time applications.

2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2022) (2022)

添加到收藏夹

Article Mathematics, Interdisciplinary Applications

Research on Surface Defect Detection of Rare-Earth Magnetic Materials Based on Improved SSD

Bin Zhang et al.

Summary: The improved SSD detection method enhances the detection performance of small defects by optimizing the model structure and feature fusion strategy. With a detection speed of 55FPS and an mAP of 83.65%, it significantly improves the ability to identify surface defects of rare-earth magnetic materials.

COMPLEXITY (2021)

添加到收藏夹

Proceedings Paper Computer Science, Artificial Intelligence

Deep Occlusion-Aware Instance Segmentation with Overlapping BiLayers

Lei Ke et al.

Summary: This study introduces a novel Bilayer Convolutional Network (BCNet) that explicitly models occlusion relationship to achieve segmentation of highly-overlapping objects, demonstrating significant performance gains in heavy occlusion cases.

2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021 (2021)

添加到收藏夹

Article Robotics