stgcn

A toolbox for skeleton-based action recognition.

implementation of STGCN for traffic prediction in IJCAI2018

[PLOS ONE] Official implementation of the paper "SignFormer-GCN : Continuous Sign Language Translation Using Spatio-Temporal Graph Convolutional Networks"

Learning Unseen Emotions from Gestures via Semantically-Conditioned Zero-Shot Perception with Adversarial Autoencoders

[SHREC24] Skeleton-based Self-Supervised Learning For Dynamic Hand Gesture Recognition

A Unified Action Recognition System that allows the users to fine-tune with only a few videos to suit their unique application.

CrisisFlow is a digital twin platform that creates a real-time, 3D virtual replica of Bangalore for crisis management

AG-MAE: Anatomically Guided Spatio-Temporal Masked Auto-Encoder for Online Hand Gesture Recognition

Two-stage skeleton-based action recognition: YOLOv8 + RTMPose + ST-GCN, C++ on NVIDIA DeepStream / TensorRT

Skeleton-based Self-Supervised Feature Extraction for Improved Dynamic Hand Gesture Recognition

Analyzing and predicting the demand for bikes using a Spatio-Temporal Graph Convolutional Network (STGCN) model.

This is the project repository for the research study Human Action Recognition using BlazePose Skeleton on Spatial Temporal Graph Convolutional Networks presented by Motasem S. Alsawadi and Miguel Rio.

Edge AI-Based Real-Time Fall Detection and Alert System for Elderly People Living Alone

Vietnamese Sign Language Recognition

Compare an athlete's form against a reference video and get back the exact joint to fix. Pose-graph analysis with MediaPipe, an adaptive ST-GCN, and DTW time alignment.

CSDC 연구실 낙상 감지 모듈

4D facial dynamics analysis with Welsh language fluency detection.

An AI-driven Adaptive Traffic Signal Control System (ATSCS) that replaces static timers with dynamic green-light phases. Utilizes YOLOv8 for real-time vehicle density estimation and multi-class classification, achieving 96.4% mAP. Optimized for low-latency inference (>30 FPS) on edge devices to reduce urban congestion and commuter wait times.

Human Activity Recognition using Deep Learning for classifying human activities from wearable sensor data through preprocessing, feature extraction, and neural network-based classification.

This repository contains the architecture, training loops, and simulation data for a smart city resource optimization engine. By leveraging Recurrent Neural Networks (RNNs) and Spatio-Temporal Graph Convolutional Networks (STGCNs), this system predicts and prioritizes the deployment of urban resources during simultaneous, colliding city events.