This project is designed to display how we can utilize deep learning methods for Sports Data Analytics.

[NeurIPS 2022 Spotlight] VideoMAE for Action Detection

[NeurIPS'22] VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training

EchoCardMAE: Video Masked Auto-Encoders Customized for Echocardiography

Real-time SL detection system featuring WebSocket streaming, VideoMAE fine-tuned on WLASL, and LLM-driven semantic disambiguation. Includes a full pipeline for 3D motion capture (JSON/Three.js) and optimized inference for CUDA and DirectML.

International Contest on Illegal Waste Dumping Detection

Real‑time crime detection system using VideoMAE, Flask, async camera streaming, alert snapshots, and email notifications.

Chitr professional case-study showcase (documentation only, no source code)

A Computer Vision Project for the AAI3001 module

Efficient video action recognition using hybrid techniques: combining ORB, SIFT, and deep models like VideoMAE and (2+1)D Conv to reduce data size while maintaining performance.

Action recognition using VideoMAE, TimeSFormer and ViViT | 1,250 video clips | 93.62% Top-1 | 98.40% Top-5 | spatio-temporal localisation

Deep learning–based anomaly detection in egocentric traffic videos using ConvAE reconstruction and VideoMAE transformer classification.

Spatio-temporal action localisation in video - VideoMAE backbone (93% Top-1 on HMDB) + DETR-style decoder for per-frame actor bounding boxes on JHMDB.

Edge-Optimized CCTV Anomaly Detection using VideoMAE.

Video action recognition using VideoMAE and SlowFast with temporal modeling