The process of classifying an image based on its visual content by assigning a label or category is known as image classification. Image classification has wide range of applications in various domains, including robotics, healthcare, biomedical devices, driverless vehicles, crop disease classifications etc. Traditional techniques for image classification such as HOG, SIFT,LBT, Gabor filters, Color Histograms, Bag of visual words and Template matching were widely used before. However, these traditional techniques have been replaced with Machine Learning or Deep Learning based techniques as they can deliver better performance in large scale image classification jobs. In this paper, a deep Convolutional Neural Network (CNN) based image classification approach has been proposed. The performance of the proposed approach has been assessed on Fruits and Vegetables Image Recognition and COVID-19 Radiography datasets. It was observed that the suggested approach delivered classification accuracy of 95.34%, 93.32%, 91.56%, 86.53% and 85.74% for Fruits and Vegetables dataset with ReLU, AReLU, Leaky ReLU, SELU and GELU respectively. For COVID-19 dataset with ReLU, AReLU, Leaky ReLU, SELU and GELU the observed accuracy values were 95.65%, 92.78%, 91.21%, 83.69% and 81.36% respectively. It was also found that the use of ReLU activation function was slightly better in terms of classification accuracy.