COMPUTER VISION


by  Martyna Malecka
Oct 2 2019, 11:43am


Does it sound familiar to you? I bet it does. Those two areas have a lot in common and are used in many, many fields.

Introduction

Computer vision has expanded into the wide area of the field from recording raw data into the extraction of image patterns and information explanation. The idea behind computer vision is to combine image processing and pattern recognition. The output of this process is image understanding. In contrast to Computer Graphics, it’s the discipline of extracting information from images. The development of computer vision depends on the computer technology system, whether about image quality improvement or image recognition. Sometimes terms of computer vision and image processing are used interchangeably, but it is not the same thing.

Computer Vision

The main goal of Computer Vision is to create models, data extracts and information from images, whereas Image Processing is more about implementing computational transformations for images, for example, contrast, sharpening, filtering. Considering functionality, computer vision and human vision are the same, in the means of interpreting special data, for example, data indexed by two or more dimensions. Nevertheless, computer vision cannot replicate just like the human eye. This is because the computer vision system has limited functions and performance with comparison to the human eye. Generally, performance evaluation involves measuring some of the basic behaviors of an algorithm to achieve strength, accuracy, or extensibility to control and monitor the performance of the system.

Usage

When it comes to practice, this technology may be used in many, many fields. Depending on the application, we would concentrate on different factors. In some cases what’s most important are the costs of production, but sometimes we care more about the time of execution. Luckily, there are a lot of technologies that help us in both cases.

Azure Form Recognizer

Considering the cost as the most important factor, we can handle it with Azure Form Recognizer. It’s a cognitive service that uses machine learning to identify and extract data from documents. It outputs structured data that includes the relationships in the original file. To work with Form Recognizer, the input documents must meet these requirements: file must be JPG, PNG or PDF. What’s best are of course text-embedded PDFs as there’s no possibility of making an error when extracting data. File size must be less than 4 MB, and for images, dimensions must be fixed. For scans, forms should be high-quality. Data, of course, must contain keys and values. When we consider these conditions, the rest is intuitive. If we need an affordable, but really effective tool, Azure Form Recognizer may be the best
solution.

AWS lambda, OpenCV, Tesseract

In the second case, when the performance counts, we can create a hybrid concerning AWS lambda services, OpenCV library, and Tesseract. This combination can result in an utterly satisfying processing time per volume. AWS Lambda is an ideal computing platform for many application scenarios. When using this, the only responsibility we have is the code. OpenCV, computer vision and machine learning software library, has more than 2.500 optimized algorithms. These algorithms may be used for detection and recognition of faces, identifying objects, classifying human actions in videos, finding similar images from an image database and many more. When we join those two with Tesseract, the optical character recognition engine which allows the user to extract the embedded text from images, we have it all, a complete solution is at fingertips.
There are many different approaches to computer vision. One can say that nowadays, it’s one of the most popular issues in the IT industry. When it comes to practice, we have a lot of possibilities. Those two mentioned above are the most convenient and efficient, but we highly recommend diving into this absorbing topic to find the perfect solution for you. Or, just leave it to TDCM, and we’ll do our best to satisfy every condition requested