The paper presents a supervised learning-based algorithm for detecting the ripeness stages of amla samples. It utilizes a simple three-class classification system. The algorithm primarily relies on a convolutional neural network (CNN) model for classification, making it computationally lightweight. Notably, the analysis is conducted solely on images captured using smartphones, enhancing the portability and potential widespread adoption of the method. Fruit maturity and quality assessment play a major role in the food industry and in harvesting. The manual grading of fruits based on their maturity levels for harvesting is a tedious process. However, the emergence of deep learning techniques has opened up the ways in this direction. We examined three convolutional neural networks utilizing a transfer-learning approach for classifying the amla fruits based on their maturity level (i.e., unripe, ripe, and over-ripe). The training/testing of models was performed over the self-collected dataset of around 535 images.