No abstract

ABSTRACT:The rapid development of Computer Vision has contributed to the widening of the techniques and methods utilized by archaeologists for the digitization and reconstruction of historic objects by automating the matching of fragments, small or large. This paper proposes a novel method for the detection of conjugate fragments, based mainly on their geometry. Subsequently the application of the Fragmatch algorithm is presented, with an extensive analysis of both of its parts; the global and the partial matching of surfaces. The method proposed is based on the comparison of vectors and surfaces, performed linearly, for simplicity and speed. A series of simulations have been performed in order to test the limits of the algorithm for the noise and the accuracy of scanning, for the number of scan points, as well as for the wear of the surfaces and the diversity of shapes. Problems that have been encountered during the application of these examples are interpreted and ways of dealing with them are being proposed. In addition a practical application is presented to test the algorithm in real conditions. Finally, the key points of this work are being mentioned, followed by an analysis of the advantages and disadvantages of the proposed Fragmatch algorithm along with proposals for future work.

Systolic Array (SA) architectures are well suited for accelerating matrix multiplications through the use of a pipelined array of Processing Elements (PEs) communicating with local connections and pre-orchestrated data movements. Even though most of the dynamic power consumption in SAs is due to multiplications and additions, pipelined data movement within the SA constitutes an additional important contributor. The goal of this work is to reduce the dynamic power consumption associated with the feeding of data to the SA, by synergistically applying bus-invert coding and zero-value clock gating. By exploiting salient attributes of state-of-the-art CNNs, such as the value distribution of the weights, the proposed SA applies appropriate encoding only to the data that exhibits high switching activity. Similarly, when one of the inputs is zero, unnecessary operations are entirely skipped. This selectively targeted, application-aware encoding approach is demonstrated to reduce the dynamic power consumption of data streaming in CNN applications using Bfloat16 arithmetic by 1%-19%. This translates to an overall dynamic power reduction of 6.2%-9.4%.

No abstract

No abstract