The prevailing methodologies for gait analysis methods require expensive and large-scale motion capture equipment, which has led to increased interest in more accessible approaches using a small number of wearable sensors. This study proposes a method for real-time estimation of hip, knee, and ankle joint angles of both legs during walking using a single, relatively inexpensive Inertial Measurement Unit (IMU) attached to the pelvis. A deep learning model based on Convolutional Neural Network (CNN) was developed and trained using approximately 16 min of data collected from a single subject. Ground truth data for training was obtained from an IMU-based motion capture system. The model’s accuracy was evaluated using the Mean Absolute Error (MAE) ± Standard Deviation (SD), averaged across all estimated joint angles, resulting in 2.0 ± 2.1 degrees. Furthermore, the trained model was implemented in an iPad application, enabling real-time estimation and visualization as a 2D animation of the lower body. This implementation demonstrates potential for extensive utilization in rehabilitation settings and sport science, representing a significant step towards realizing a low-cost, location-independent, and portable gait analysis solution.