Artistic swimming is a sport that combines elements of dance, swimming, and gymnastics. Traditionally the sport is evaluated by human judges which introduces subjectivity and inconsistency to the evaluation metrics. This paper proposes a machine learning-based approach to automate the evaluation process of artistic swimming routines. The model uses a convolutional recurrent neural network (CRNN) architecture built on Keras and TensorFlow. It extracts features from video frames using convolutional layers and the recurrent layers help detect the sequence of postures making up each routine. The score is then calculated with OpenCV using key point detection. The overall framework helps input videos, classify posture, and detect sequences and computes performance score to give a more objective and consistent result along with valuable feedback to the athlete based on the scores generated by the ML model.