Abstract Social progress and technological development have brought sweeping changes to the field of art education. The innovative development of piano art instruction, as a part of art education, has a significant impact on students and the academic world. The study proposes a deep learning-based model for piano art instruction, taking into account the current state of piano art instruction, and investigates the impact of this model on students’ piano playing skills and appreciation of music art. The experimental group’s students improved their playing skills in the eight dimensions of fingering, timbre, pitch, rhythm, musical expression, emotional output, body language, and stage image. Except for timbre and stage image, the p-values of the experimental group and the control group in the other six dimensions of piano playing skills were less than 0.05, which effectively improved. In terms of music art appreciation, the total number of excellences in the three dimensions of aesthetic perception, artistic performance, and cultural understanding was 118, 131, and 126, respectively, and the total number of excellences in the experimental group was 32, 45, and 39. Obviously, piano art instruction based on deep learning has a positive impact on the students’ piano-playing skills and ability to appreciate music art.