Intelligent Classification of Electrical Equipment Using Machine Learning

The report presents a multi-class classifier designed and tested for classes LIGHT SWITCH, POWER OUTLET and DB BOARD. It has chosen the number of images in the dataset at 117 where the numbers have been divided between training set and test set at 90 training images and 27 test images. This model showed phenomenal ability, with a pretty good performance attained accuracy on the test set, which manifests its success for that particular task. Therefore, this project further confirms the capability of the platform of Teachable Machine in the design of an image classification model even for less professionalized machine learning experts.

While creating such an architecture model, the critical methodology was using a CNN in such a way that it learns the key features present in the input images. Most importantly, enforcing high accuracy levels will be the hierarchical representations of visual data that a CNN can learn. Techniques of preprocessing data have ensured uniformity and consistency in input data, which further enhanced the performance by the model.