Back to selected workPROJECT / MOTIONSENSE

Computer vision · Research

MotionSense

Human activity recognition from video, using a MobileNetV3 feature extractor and temporal LSTM model.

Technology
PyTorchMobileNetV3LSTMOpenCVStreamlit
Project notesFeatured work · Updated Nov 2026
MotionSense studies human activity recognition from video using a MobileNetV3-Large frame encoder and an LSTM over short frame windows. Its Streamlit application is the inference surface for video upload and results; model training stays in the original repository.
Interactive demonstrationThe project entry is live. Its hosted demonstration URL can be added from the project dashboard when ready.