Python action sequences for next-method-call token prediction (Internetware 2022): paper, AST extraction demo, method diagrams, results, and reproducibility guide.
-
Updated
Oct 19, 2026 - Python
Python action sequences for next-method-call token prediction (Internetware 2022): paper, AST extraction demo, method diagrams, results, and reproducibility guide.
For HAR, our novel HDAD (IGAV) dataset was constructed by performing 4 dynamic and 3 static activities with the accelerometer and gyroscope sensors of the IOS smart phone in two different positions for a total of 15 seconds. Mentioned activities were collected in real time by placing them on the waist of a total of 10 volunteers.
Vision-Guided Smart Home Control is an innovative project that combines the power of computer vision with the connectivity of IoT to create a seamless human activity recognition system. Using OpenCV for accurate motion detection and pySerial for IoT integration, ActionSense transforms how environments respond to human actions, enhancing automation.
An late fusion architecture between I3D and Xception for Human Activity Recognition
Human Activity Recogition GUI
This repo is to investigate Neural Networks for recognition of human activities.
To associate your repository with the humanactivityrecognition topic, visit your repo's landing page and select "manage topics."