Data Extraction of Circular-Shaped and Grid-like Chart Images

Abstract

Chart data extraction is a crucial research field in recovering information from chart images. With the recent rise in image processing and computer vision algorithms, researchers presented various approaches to tackle this problem. Nevertheless, most of them use different datasets, often not publicly available to the research community. Therefore, the main focus of this research was to create a chart data extraction algorithm for circular-shaped and grid-like chart types, which will accelerate research in this field and allow uniform result comparison. A large-scale dataset is provided containing 120,000 chart images organized into 20 categories, with corresponding ground truth for each image. Through the undertaken extensive research and to the best of our knowledge, no other author reports the chart data extraction of the sunburst diagrams, heatmaps, and waffle charts. In this research, a new, fully automatic low-level algorithm is also presented that uses a raster image as input and generates an object-oriented structure of the chart of that image. The main novelty of the proposed approach is in chart processing on binary images instead of commonly used pixel counting techniques. The experiments were performed with a synthetic dataset and with real-world chart images. The obtained results demonstrate two things: First, a low-level bottom-up approach can be shared among different chart types. Second, the proposed algorithm achieves superior results on a synthetic dataset. The achieved average data extraction accuracy on the synthetic dataset can be considered state-of-the-art within multiple error rate groups.

Validation of Glycemic Index Values Represented as Bar Graphs in Scholarly Articles for Clinical Nutrition Application

“…Otherwise, we solely encountered GI values that were represented as vertical bars from the 19 scholarly articles analysed in this research. Typically, bar graphs could be divided into basic, grouped, or stacked bars represented either vertically or horizontally as suggested by Bajic et al, and Mishra et al, in their respective research [16,49]. Based on this premise, we deduce that ImageJ could also be potentially utilized for data mining in linear plot graphs, box plots, 3-Dimensional (3D) bar graphs that are represented as single, grouped, or stacked graphs.…”

“…In general, according to Bajic et al, data represented as figures could be either automatically or interactively extracted [16]. These could be carried out via photo editing programs, image processing programs or custom coded algorithms.…”

Deep learning of electrochemical CO2 conversion literature reveals research trends and directions

“…The tabular data from the tables of a paper could be handled using data analysis libraries including Pandas; 89 however, it is important to standardize the tabular data because the formats of the table are much different across papers and journals. For gures, numerous vision techniques are rapidly developing which are applicable to a graph 90 and microscopy images, 91,92 which were proven to be very effective to extract the data from the gures in papers. Lastly, we observed that the NER performance is lower for catalyst name entities than for other entities, probably because researchers describe catalyst names in various forms in different papers.…”

Automatic Pipeline for Information of Curve Graphs in Papers Based on Deep Learning

“…Zhou et al devised a method to generate textual descriptions for graphs that summarize information about the value relationships [32]. Researchers have also investigated automated data extraction from simple numerical charts, such as bar charts and pie charts [10,33,34]. For curve graphs, Figureseer can analyze the content on the graph while extracting the numerical value, including the association information of the legend entries [35].…”