Applying Deep Learning Approaches to Volleyball Data

[ICCV 2021] A new codebase containing various methods for Group Activity Recognition. Paper title: Spatio-Temporal Dynamic Inference Network for Group Activity Recognition.

A curated list of Group Activity (a.k.a., Collective Activity) recognition papers, codes, datasets and other resources.

The official repository of the paper "SGA-INTERACT: A 3D Skeleton-based Benchmark for Group Activity Understanding in Modern Basketball Tactic"

[NeurIPS 2024] CHASE: Learning Convex Hull Adaptive Shift for Skeleton-based Multi-Entity Action Recognition

Panoramic Human Activity Recognition, in ECCV 2022.

A modern implementation of CVPR 2016 paper: "A Hierarchical Deep Temporal Model for Group Activity Recognition."

Modern implementation of CVPR16 "Hierarchical Deep Temporal Models for Group Activity Recognition" paper

Enhanced adaptation of MP-GCN for recognizing playground activities using only 2D skeletons and geometric cues. It builds spatio-temporal graphs that capture interactions among people and playground structures, supported by a streamlined pipeline for pose detection, tracking, graph construction, data stratification, and multi-stream training

A new paper implementaion "Hierarchical Relational Networks for Group Activity Recognition and Retrieval"

▎ PyTorch implementation of group activity recognition on the CVPR 2016 volleyball dataset — 8 baselines from single-frame CNNs to hierarchical LSTMs, with a unified data loader, Hydra configs, and Kaggle-ready training.

PyTorch implementation of a Relational Attention-Based Model for Group Activity Recognition — inspired by "A Hierarchical Deep Temporal Model for Group Activity Recognition (arXiv:1511.06040v2)"

2020 CVPR Workshop Paper -- RIT-18: A Novel Dataset for Compositional Group Activity Understanding

Automate gameplay in Volleyball Legends with aim assistance and utility tools for web browsers.

A modern PyTorch implementation of the CVPR 2016 paper "Hierarchical Deep Temporal Models for Group Activity Recognition"

A modern PyTorch implementation of hierarchical deep temporal models for group activity recognition

A modern implementation of ECCV 2018 paper: "Hierarchical Relational Networks for Group Activity Recognition and Retrieval".

Implementation of "Hierarchical Relational Networks for Group Activity Recognition" (ECCV 2018) with ResNet50, PyTorch Geometric relational layers, graph attention and temporal LSTM variants — outperforming the original paper across all baselines on the volleyball dataset. 🏐

Modern & enhanced implementation of Hierarchical Deep Temporal Models for Group Activity Recognition. Achieves 90.2% accuracy on the Volleyball Dataset using Global-Local Feature Fusion.

This project implements a hierarchical deep learning approach for recognizing group activities in videos. The core idea is to model both individual person actions and group-level activities using temporal dynamics captured by LSTM networks.