Abstract. I/O virtualization performance is an important problem in KVM.In this paper, we evaluate KVM I/O performance and propose several optimizations for improvement. First, we reduce VM Exits by merging successive I/O instructions and decreasing the frequency of timer interrupt. Second, we simplify the Guest OS by removing redundant operations when the guest OS operates in a virtual environment. We eliminate the operations that are useless in the virtual environment and bypass the I/O scheduling in the Guest OS whose results will be rescheduled in the Host OS. We also change NIC driver's configuration in Guest OS to adapt the virtual environment for better performance.

No abstract

No abstract

Localizing text instances in natural scenes is regarded as a fundamental challenge in computer vision. Nevertheless, owing to the extremely varied aspect ratios and scales of text instances in real scenes, most conventional text detectors suffer from the sub-text problem that only localizes the fragments of text instance (i.e., sub-texts). In this work, we quantitatively analyze the sub-text problem and present a simple yet effective design, COntrastive RElation (CORE) module, to mitigate that issue. CORE first leverages a vanilla relation block to model the relations among all text proposals (subtexts of multiple text instances) and further enhances relational reasoning via instance-level sub-text discrimination in a contrastive manner. Such way naturally learns instance-aware representations of text proposals and thus facilitates scene text detection. We integrate the CORE module into a twostage text detector of Mask R-CNN and devise our text detector CORE-Text. Extensive experiments on four benchmarks demonstrate the superiority of CORE-Text. Code is available: https://github.com/jylins/CORE-Text.

No abstract