In today’s manufacturing world, it is very crucial to keep an eye on the state of the manufacturing environment. Quality control (QC) procedures can also be used to continuously evaluate a product's health over the course of its production lifespan. Several factors affect the visual inspection procedure, leading to an industry-wide inspection accuracy of about 80%. To make the inspection process more user-friendly, the approach combines a customized convolution neural network (CNN) for inspection and a computer program that can be installed on the shop floor. Here, the casting product dataset is the input. The dataset comes from the Technocast pilot. The conventional quality inspection methods consume more time and labor cost and also reduce the accuracy level. In the proposed methodology, the deep learning-based quality inspection method is proposed in order to find the quality of product as defective or non-defective. The various deep learning-based CNN architectures are used for predicting the inspection and its results of training and validation accuracy and losses are discussed.