Deep convolutional neural networks (DCNN) have made significant progress in a wide range of applications in recent years, which include image identification, audio recognition, and translation of machine information. These tasks assist machine intelligence in a variety of ways. However, because of the large number of parameters, float manipulations and conversion of the machine terminal remains difficult. An optimisation process is initiated to address this issue by adjusting the neural network's convolution characteristics, which minimises information loss and enhances its overall performance. Minimisation of the convolution function addresses the optimisation issues. Initially, batch normalisation is completed, and instead of lowering neighbourhood values, a full feature map is minimised to a single value using the global pooling approach. Traditional convolution is split into depth and pointwise to decrease the model size and calculations. The optimised convolution-based DCNN's performance is evaluated with the assistance of accuracy and occurrence of error. The optimised DCNN is compared with the existing state-of-the-art techniques, and the optimised DCNN outperforms the existing technique.