Modern object detectors usually suffer from low accuracy issues, as foregrounds always drown in tons of backgrounds and become hard examples during training. Compared with those proposal-based ones, real-time detector...
详细信息
Modern object detectors usually suffer from low accuracy issues, as foregrounds always drown in tons of back-grounds and become hard examples during training. Compared with those proposal-based ones, real-time detecto...
详细信息
Modern object detectors usually suffer from low accuracy issues, as foregrounds always drown in tons of back-grounds and become hard examples during training. Compared with those proposal-based ones, real-time detectors are in far more serious trouble since they renounce the use of region-proposing stage which is used to filter a majority of back-grounds for achieving real-time rates. Though foregrounds as hard examples are in urgent need of being mined from tons of backgrounds, a considerable number of state-of-the-art real-time detectors, like YOLO series, have yet to profit from existing hard example mining methods, as using these methods need detectors fit series of prerequisites. In this paper, we propose a general hard example mining method named Loss Rank Mining (LRM) to fill the gap. LRM is a general method for real-time detectors, as it utilizes the final feature map which exists in all real-time detectors to mine hard examples. By using LRM, some elements representing easy examples in final feature map are filtered and detectors are forced to concentrate on hard examples during training. Extensive experiments validate the effectiveness of our method. With our method, the improvements of YOLOv2 detector on auto-driving related dataset KITTI and more general dataset PASCAL VOC are over 5% and 2% mAP, respectively. In addition, LRM is the first hard example mining strategy which could fit YOLOv2 perfectly and make it better applied in series of real scenarios where both real-time rates and accurate detection are strongly demanded.
Hardware-based middleboxes are ubiquitous in computer networks, which usually incur high deployment and management expenses. A recently arsing trend aims to address those problems by outsourcing the functions of tradi...
详细信息
New non-volatile memory (e.g., phase-change memory) provides fast access, large capacity, byteaddressability, and non-volatility features. These features, fast-byte-persistency, will bring new opportunities to fault...
详细信息
New non-volatile memory (e.g., phase-change memory) provides fast access, large capacity, byteaddressability, and non-volatility features. These features, fast-byte-persistency, will bring new opportunities to fault tolerance. We propose a fine-grained checkpoint based on non-volatile memory. We extend the current virtual memory manager to manage non-volatile memory, and design a persistent heap with support for fast allocation and checkpointing of persistent objects. To achieve a fine-grained checkpoint, we scatter objects across virtual pages and rely on hardware page-protection to monitor the modifications. In our system, two objects in different virtual pages may reside on the same physical page. Modifying one object would not interfere with the other object. This allows us to monitor and checkpoint objects smaller than 4096 bytes in a fine-grained way. Compared with previous page-grained based checkpoint mechanisms, our new checkpoint method can greatly reduce the data copied at checkpoint time and better leverage the limited bandwidth of non-volatile memory.
distributed software systems are becoming more and more complex *** is easy to find a huge amount of computing nodes in a nationwide or global information *** example,We Chat(Wei Xin),a well-known mobile application i...
详细信息
distributed software systems are becoming more and more complex *** is easy to find a huge amount of computing nodes in a nationwide or global information *** example,We Chat(Wei Xin),a well-known mobile application in China,has reached a record of 650 million monthly active users in the third quarter of *** the same time,researchers are starting to talk about software systems which have billions of lines of codes[1]or can last one hundred years.
Audio feature extraction is a very important technique in the field of sound processing. It extremely impacts the effectiveness and correctness of sound recognition, sound verification, etc. It is a computation intens...
详细信息
Audio feature extraction is a very important technique in the field of sound processing. It extremely impacts the effectiveness and correctness of sound recognition, sound verification, etc. It is a computation intensive stage in the whole sound recognition process, which is a challenging for acceleration. In this paper, a coarse-grained parallel feature extraction algorithm for high throughput of audio slices is proposed to improve the efficiency of audio feature extraction. Three typical audio feature extraction algorithms, Mel Frequency Cepstrum Coefficients(MFCC), Spectrogram image features(SIF), Octave-Based Spectral Contrast, are chosen to parallelize. Experiments results on different platforms show that the speedup of accelerated audio feature extraction is up to 17.23 on the platform with 16 cores 32 threads.
Zero pronoun resolution is very important in natural language processing. Identification of zero pronoun is the fundamental task of its resolution. Existing feature engineering based identification approaches are unsu...
详细信息
ISBN:
(纸本)9781510845541
Zero pronoun resolution is very important in natural language processing. Identification of zero pronoun is the fundamental task of its resolution. Existing feature engineering based identification approaches are unsuitable for big data applications due to labor-intensive work. Furthermore, as extracted from parse trees which are not unique for a certain sentence, features may be improper for zero pronoun identification. In this paper, we constructed a two-layer stacked bidirectional LSTM model to tackle identification of zero pronoun. To extract semantic knowledge, the first layer obtains the structure information of the sentence, and the second layer combines the part-of-speech information with obtained structure information. The results in two different kinds of experimental environment show that, our approach significantly outperforms the state-of-the-art method with an absolute improvement of 4.3% and 20.3% F-score in Onto Notes 5.0 corpus respectively.
In this paper, we make a research on a widely-used SAT solver, Minisat, aiming to improve its performance using coarse-grained parallel method on multi-core and multi-platform. Firstly, we parallel the Minisat by mean...
详细信息
In this paper, we make a research on a widely-used SAT solver, Minisat, aiming to improve its performance using coarse-grained parallel method on multi-core and multi-platform. Firstly, we parallel the Minisat by mean of Open MP and test its performance with different threads by running a test set consisting of 2000 SAT problems on an X86 computer. Besides, a scheduling strategy with time sequence is added to the process and achieves a better speed-up ratio. Then, we move the algorithm to an ARM computer and repeat the same process, finding that the performance of Minisat on X86 is better than that on ARM, but ARM platform has a better scale effect than X86 platform when running at full load and is able to perform better than X86 when they have the same hardware configuration.
Audio tagging aims to infer descriptive labels from audio clips. Audio tagging is challenging due to the limited size of data and noisy labels. In this paper, we describe our solution for the DCASE 2018 Task 2 general...
详细信息
Dangling pointer error is pervasive in C/C++ programs and it is very hard to detect. This paper introduces an efficient detector to detect dangling pointer error in C/C++ programs. By selectively leave some memory acc...
详细信息
暂无评论