Figure 1.1 demonstrates contrasting feature detection which is critical in creating an even distribution of data. In other words, image variance and randomness is minimized with the implementation of edge-based detection as it blends and identifies local regions of the image. Leading up to preprocessing test trials, the accepted conjecture suggests that edge-based detection will maximize accuracy of data by easing the machine learning process. Moreover, this will decrease time-consuming and inefficient neural network iterations leading to scalability and faster learning paces. By using borderline/edge-based detection, critical regions of the image(s) are revealed to the API training process, advancing the working neural network algorithms.
In addition to an edge-based detection method, gamma image preprocessing was considered in the criteria of the project. Gamma detection is comprised of a simple input-output gamma curve. More importantly, it is the pixel value relationship between the input and output. It is identified with a reciprocal value that becomes increasingly proportional. The resulting image highlights and partly brightens colorated features, etc. Purposed as a preprocessing technique, gamma detection reveals darkened local regions of the image to the eventual User Interface (UI) in the application. A processing example of the gamma image argument is demonstrated below (Figure 1.2) with a gamma intensity of ~7.12.