This project involves the development of a Convolutional Neural Network (CNN) for image classification. The entire project is conducted using Google Colaboratory, an online source IDE. The project's dataset consists of images of an object taken from various directions, including top, bottom right, bottom left, top right, and top left. These videos were converted into images and organized into folders, with each folder containing around 1500 images, resulting in a total of 7500 images.
The data underwent preprocessing to reduce the image size and facilitate processing on the GPU to alleviate the computational load. One major challenge faced during the project was the variation in angles of the object. To address this issue, the developer implemented code to adjust the angles of the images and saved the modified data to Google Drive.
The dataset was then labeled, with the images being represented as 'x' (features) and their corresponding labels as 'y'. The labeled data was fed into the CNN algorithm, which was trained to classify the images into different categories corresponding to the object's directions. The project achieved an impressive accuracy result of 0.9975, indicating that the CNN model performed exceptionally well in classifying the images based on their angles.