Fish have always played a vital role in human Life not only in the field of aquaculture but also in medical science, environment conservation, and many other fields but as there is a huge range of species of fish which makes it really challenging to correctly classify them and people remain unaware of there types and uses. So, we used MobilenetV2 a CNN model to overcome these problems. Pre-trained MobilenetV2 model is used in this research to test its ability in low-end devices such as Phones and other embedded devices as MobilenetV2 is a lightweight CNN model. Fish classification is crucial in today’s life for conserving native fish species and ecosystems. For instance, some of the fish have a medical property and are used in traditional medicine and other medical treatments, correct classification of these fish also helps in identifying fish for sustaining Aquaculture and studying marine Ecology. In this research, we have worked on the Fish4Knowledge dataset having 27000 images with 23 categories of fishes divided into a ratio of 80% training data, 10% validation data, and 10% test data. After the training and testing, the highest accuracy achieved was 95.20% for training and 98% for testing We conducted several experiments and tuning to achieve this using the MobilenetV2 model.