Plug-and-Play Few-shot Object Detection with Meta Strategy and Explicit Localization Inference

Aiming at recognizing and localizing the object of novel categories by a few reference samples, few-shot object detection is a quite challenging task. Previous works often depend on the fine-tuning process to transfer their model to the novel category and rarely consider the defect of fine-tuning, resulting in many drawbacks. For example, these methods are far from satisfying in the low-shot or episode-based scenarios since the fine-tuning process in object detection requires much time and high-shot support data. To this end, this paper proposes a plug-and-play few-shot object detection (PnP-FSOD) framework that can accurately and directly detect the objects of novel categories without the fine-tuning process. To accomplish the objective, the PnP-FSOD framework contains two parallel techniques to address the core challenges in the few-shot learning, i.e., across-category task and few-annotation support. Concretely, we first propose two simple but effective meta strategies for the box classifier and RPN module to enable the across-category object detection without fine-tuning. Then, we introduce two explicit inferences into the localization process to reduce its dependence on the annotated data, including explicit localization score and semi-explicit box regression. In addition to the PnP-FSOD framework, we propose a novel one-step tuning method that can avoid the defects in fine-tuning. It is noteworthy that the proposed techniques and tuning method are based on the general object detector without other prior methods, so they are easily compatible with the existing FSOD methods. Extensive experiments show that the PnP-FSOD framework has achieved the state-of-the-art few-shot object detection performance without any tuning method. After applying the one-step tuning method, it further shows a significant lead in both efficiency, precision, and recall, under varied few-shot evaluation protocols.

Plug-and-Play Few-shot Object Detection with Meta Strategy and Explicit Localization Inference

Aiming at recognizing and localizing the object of novel categories by a few reference samples, few-shot object detection is a quite challenging task. Previous works often depend on the fine-tuning process to transfer their model to the novel category and rarely consider the defect of fine-tuning, resulting in many drawbacks. For example, these methods are far from satisfying in the low-shot or episode-based scenarios since the fine-tuning process in object detection requires much time and high-shot support data. To this end, this paper proposes a plug-and-play few-shot object detection (PnP-FSOD) framework that can accurately and directly detect the objects of novel categories without the fine-tuning process. To accomplish the objective, the PnP-FSOD framework contains two parallel techniques to address the core challenges in the few-shot learning, i.e., across-category task and few-annotation support. Concretely, we first propose two simple but effective meta strategies for the box classifier and RPN module to enable the across-category object detection without fine-tuning. Then, we introduce two explicit inferences into the localization process to reduce its dependence on the annotated data, including explicit localization score and semi-explicit box regression. In addition to the PnP-FSOD framework, we propose a novel one-step tuning method that can avoid the defects in fine-tuning. It is noteworthy that the proposed techniques and tuning method are based on the general object detector without other prior methods, so they are easily compatible with the existing FSOD methods. Extensive experiments show that the PnP-FSOD framework has achieved the state-of-the-art few-shot object detection performance without any tuning method. After applying the one-step tuning method, it further shows a significant lead in both efficiency, precision, and recall, under varied few-shot evaluation protocols.

Design and implementation of pipelined and parallel AES encryption systems using FPGA

<p><span>The information security is one of the most important issues in the design of any communication network.One of the most common encryption algorithms is the Advanced Encryption Standard (AES).The main problem facing the AES algorithm is the high time consumption due to the large number of rounds used for performing the encryption operation. The more time the encryption system consumes to encrypt the data, the more chances the hackers have to break the system.This paper presents two effective algorithms that can be used to solve the mentioned problem and to achieve an effective processing time reduction using pipelined and parallel techniques to perform the encryption steps. These algorithms are based on using certain techniques to make the system able to encrypt many different states (the data will be encrypted) in the same time with no necessity to wait for the previous encryption operation to be completed. These two algorithms are very effective especially for big data size. This paper describes in detail the AES encryption system algorithm and a detailed explanation for the proposed algorithms. Moreover, the research shows the implementation of the three algorithms: the traditional, the pipelined, and the parallel algorithms, and finally a comparison between them.</span></p>

Analogue x digital: Parallel techniques for design learning

With urban space under the ubiquitous scrutiny of digital visioning technologies, the city is now imaged by means of a pixel grid containing ephemeral, qualitative data presented as colour, brightness and shape. Unlike traditional analogue pictorial representational modes, the digital image is a highly transformable mechanism with an unstable distribution of data across its pixel array. As a consequence, representation in the form of spatial abstraction demands not only a new approach to the learning and implementation of traditional disciplinary drawing practice, but a rethinking of the alignment and cooperative nature of analogue and digital drawing models when applied to effective design development. In a pedagogical context, the transition of spatial representation between analogue and digital modes has profound implications for how the student connects seminal drawing and design processes to both the sensorial realm and the physical experience of lived space. This paper therefore explores the enhancement of tertiary learning in digital and abstract literacy through new drawing techniques. Underpinned by a new relationship between representation and envisioned physical space, the techniques are applied within learning environments in parallel with existing analogue pictorial procedures. By building curriculum for foundational students that provides a framework of linked spatial experiences aligned across analogue and digital domains and coupled with tasks focused on the development of conceptual thinking, it proposes increased student success in future studios and professional practice.

Non-Stationary Acceleration Strategies for PageRank Computing

In this work, a non-stationary technique based on the Power method for accelerating the parallel computation of the PageRank vector is proposed and its theoretical convergence analyzed. This iterative non-stationary model, which uses the eigenvector formulation of the PageRank problem, reduces the needed computations for obtaining the PageRank vector by eliminating synchronization points among processes, in such a way that, at each iteration of the Power method, the block of iterate vector assigned to each process can be locally updated more than once, before performing a global synchronization. The parallel implementation of several strategies combining this novel non-stationary approach and the extrapolation methods has been developed using hybrid MPI/OpenMP programming. The experiments have been carried out on a cluster made up of 12 nodes, each one equipped with two Intel Xeon hexacore processors. The behaviour of the proposed parallel algorithms has been studied with realistic datasets, highlighting their performance compared with other parallel techniques for solving the PageRank problem. Concretely, the experimental results show a time reduction of up to 58.4 % in relation to the parallel Power method, when a small number of local updates is performed before each global synchronization, outperforming both the two-stage algorithms and the extrapolation algorithms, more sharply as the number of processes increases.

Kualitas radiograf periapikal dengan teknik paralel di Rumah Sakit Gigi dan Mulut Universitas PadjadjaranQuality of periapical radiographs with parallel techniques in Universitas Padjadjaran Dental Hospital

Pendahuluan: Radiografi sering digunakan untuk pemeriksaan penunjang utama dalam menegakkan diagnosis kelainan periapikal dan yang paling sering digunakan adalah teknik intraoral periapikal. Menegakkan diagnosis secara tepat harus memperhatikan kualitas radiograf, karena digunakan sebagai alat bantu dalam penegakan diagnosis, penentuan rencana perawatan, dan evaluasi pasca perawatan. Teknik periapikal paralel menjadi pilihan terbaik karena dapat menghasilkan radiograf dengan distorsi minimal dan akurasi linier yang lebih akurat. Tujuan penelitian ini adalah untuk melihat kualitas radiograf perapikal dengan teknik paralel di Rumah Sakit Gigi dan Mulut Universitas Padjajaran. Metode: Jenis penelitian ini adalah deskriptif. Populasi penelitian adalah seluruh radiograf dengan teknik paralel pada tahun 2018. Sampel penelitian ditentukan dengan metode consecutive sampling pada bulan Oktober-Desember 2018, sehingga didapatkan sebanyak 52 sampel radiograf periapikal dengan teknik paralel. Hasil: Terdapat 27 radiograf periapikal (56%) memiliki kriteria penilaian sempurna, 22 radiograf periapikal (42%) memiliki kriteria penilaian dapat diterima secara diagnostik dengan terdapat beberapa kesalahan, dan 3 radiograf periapikal (6%) memiliki penilaian tidak baik sehingga tidak dapat diterima secara diagnostik berdasarkan National Radiological Protection Board serta pemenuhan kriteria kualitas paling banyak adalah kontras dan distorsi ukuran. Simpulan: Kualitas radiografi periapikal dengan teknik paralel di RSGM UNPAD secara umum dapat diterima secara diagnostik dengan penilaian kualitas berada pada rating 1 yaitu, tidak adanya kesalahan dari persiapan pasien, eksposure, posisi film holder, dan pengolahan film, serta dikatakan sempurna secara visual.Kata kunci: Kualitas radiograf, teknik periapikal paralel ABSTRACTIntroduction: Radiographs are often used for primary investigations in the diagnosis of periapical abnormalities, and the most frequently used is the intraoral periapical technique. Establishing appropriate diagnosis emphasise the radiograph quality, because it is used as a tool in establishing diagnoses, determining treatment plans, and post-treatment evaluation. The parallel periapical technique is the best choice because they can produce radiographs with minimal distortion and more accurate linear accuracy. The purpose of this study was to observe the quality of parallel imaging radiographs in Universitas Padjadjaran Dental Hospital. Methods: The type of research was descriptive. The study population was all radiographs with parallel techniques in 2018. The study sample was determined by the consecutive sampling method in October-December 2018, resulting in 52 samples of periapical radiographs with parallel technique. Results: There were 27 periapical radiographs (56%) with perfect assessment criteria, 22 periapical radiographs (42%) had diagnostic criteria of acceptable with several errors, and 3 periapical radiographs (6%) had poor assessment, so they were not acceptable for diagnostic use based on the National Radiological Protection Board and the fulfilment of the most quality criteria was the contrast and size distortion. Conclusion: The periapical radiographs quality with parallel techniques in Universitas Padjadjaran Dental Hospital can be generally accepted diagnostically with the assessment of quality at rating 1, namely, the absence of errors from patient’s preparation, exposure, film holder position, and film processing, and is said to be visually perfect.Keywords: Radiograph quality, parallel periapical technique