China
Africa
Explore chapters and articles related to this topic
Data Formats
Published in Praveen Kumar, Jay Alameda, Peter Bajcsy, Mike Folk, Momcilo Markus, Hydroinformatics: Data Integrative Approaches in Computation, Analysis, and Modeling, 2005
We have seen how the choice of a file structure should correspond to the types of uses that are anticipated, and the kind of data being represented. For instance, we have seen that fixed length record files and indexed files can improve our ability to access data directly and to randomly search for records. Because values in binary files are of a fixed size, binary files are more easily organized into fixed length records than are text files. Text files can be organized with fixed length records, but the result can be that records are much larger than they would otherwise need to be. Similarly, indexed files can be either binary or text, but the resulting text files can be much larger and not as easily converted to a form that computers easily deal with. This is particularly the case for complex index structures, such as search trees.
Management of LiDAR Data
Published in Jie Shan, Charles K. Toth, Topographic Laser Ranging and Scanning, 2017
Elevation data are commonly encoded as either binary or American Standard Code for Information Exchange (ASCII). Binary encoding is compact, using the minimum file space possible for each data item. For example, if the height values (z) are stored as 32-bit integers, then this is all the space that is occupied per height value (4 bytes per record). Generally, binary files cannot be read without agreement on both the writer's and reader's part of the specific binary format (i.e., 32-bit integer, signed, little-endian format).
Data Storage
Published in Chandrasekar Vuppalapati, Building Enterprise IoT Applications, 2019
Binary files3 also usually have faster read and write times than text files, because a binary image of the record is stored directly from memory to disk (or vice versa). In a text file, everything has to be converted back and forth to text, and this takes time.
Counting of exposed aggregate number on pavement surface based on computer vision technique
Published in Road Materials and Pavement Design, 2023
Lyhour Chhay, Lae-Jeong Park, Young Kyu Kim, Seung Woo Lee
Initially, the dataset generated through LabelImg was converted into a comma separated value (CSV) file format to avoid the complexity of labelling numerous XML files. Because working directly with the image dataset and labelling generates a large input, the binary file format was selected to store the data. Therefore, the dataset was merged and converted into a record file format called TFRecord. TFRecord can easily store a large dataset during the training process. After generating TFRecord, we created a label map namely ‘Aggregate' as the annotation process then conducted the configuration for training. In the configuration file, we adopted the original configuration file from Faster R-CNN inception v2 in which the learning rate was set to 0.002 in the initialisation stage and 0.0002 after 9000 interactions; other hyper-parameters are shown in Table 1. The model was trained on Windows 10 Education 64 bit with the specification Intel R Xeon(R) CPU E5-2630 v4 @2.20GHz×20 processor, 64 GB RAM, GPU NVIDIA Quadro P2000 5GB DDR6.
On the Effectiveness of Image Processing Based Malware Detection Techniques
Published in Cybernetics and Systems, 2022
Markov Plot: Another image representation technique based on the first order Markov model is suggested in [17, 16]. The method starts with constructing a weighted directed graph G = < V, E >, with 256 vertices. Each vertex , indicates possible byte values from 0 to 255, and each directed edge ex, corresponds to the transition from byte value x to y. Initially, the weight of all the edges is set to 0. For the conversion process, the binary file is read by byte from start to end. Whenever there is a transition from byte value x to y, the graph draws a directional edge from x to y and increases the edge weight by 1. Refer w(ex,) as the total number of edges from x to y. To calculate the transition probability Pr(xy|ex) (probability of occurring byte value y just after x), each edge weight is divided by the sum of all outbound edge weights from the source node (here x) as shown in Equation (1).
Cloud manufacturing architecture for part quality assessment
Published in Cogent Engineering, 2020
Alessandra Caggiano, Tiziana Segreto, Roberto Teti
The UT node in the fog layer received from the UT NDI robotic system data related to the UT scan parameters and the digitised UT waveforms, with file extension driven by the scan software utilized at the device layer and requiring conversion into a format readable by the subsequent processing module at the cloud layer. Data pre-processing at the UT node in the fog layer consisted of the organization of the UT waveforms as a 3D matrix, where the rows and columns correspond to the x and y coordinates of the scan points while the third dimension includes the full-digitized UT waveform for each scan point. These pre-processed data were saved under binary file format and sent via HTTPS data communication protocol to the cloud server.