Rice is a popular Asian food grain which grows in natural conditions. Different diseases can negatively impact its production quality and quantity. In recent years, machine learning techniques have proven their utility in rice plant disease detection. The ML-based methods require a set of predetermined examples for learning. The dataset preparation and annotation require human effort and time. This paper proposes a deep learning approach to accurately detect rice plant disease using raw images. The contributed model uses the transfer learning concept to capture image features and classify them. The model has been developed using VGG-16 and deep convolutional neural network (CNN). The model has also been compared with the traditional image classification techniques based on features like Sobel and k-means segmentation. The Mendeley dataset was used to perform training and validation of the models. The experimental result demonstrates that the sequential CNN can provide 76% accurate classification. Additionally, the 2D-CNN with the same dataset provides 96% accuracy. Similarly, the traditional feature Sobel and 2D-CNN deliver 81% accuracy. Furthermore, the combination of k-means and 2D-CNN offers 79% accuracy. Finally, the proposed technique based on VGG-16 and 2D-CNN provides 99% accuracy with very few epochs.