Identifying and classifying the species of pollen grains is an intricate and time-consuming process. To simplify this process, various machine learning and deep learning models were created, but most of them are computationally expensive. This motivated the authors to carry out a study with pollen grains collected from the southern part of India and categorize them into three species using deep convolutional neural networks that are not computationally expensive. The study focuses on evaluating and comparing several well-known Convolutional Neural Network architectures-GoogLeNet, LeNet, Residual Neural Network (ResNet), AlexNet, and Visual Geometry Group 16 (VGG16) to determine their effectiveness in classifying the pollen images into their respective classes. It was proved that the results of LeNet outpowered the other networks with a mean accuracy, loss and F1 score of 99.22%, 1.65% and 99.02%, respectively. These results highlights the algorithm’s effectiveness in accurately classifying the diverse plant species based on the input images.