Cricket Analyzer Application (Training, Inference and Complete DataSet)

## Cricket_Analyzer — Full Dataset (~20 GB) for Real-time Cricket Ball Speed Estimation & Trajectory Tracking This Zenodo record hosts the **full Cricket_Analyzer dataset (~20 GB)** used in our study on **vision-based, real-time cricket ball detection, trajectory tracking, and ball speed estimation**. It is deposited to support **PLOS ONE reproducibility and data availability** requirements by providing the data needed to replicate all findings reported in the manuscript. ### Related Code + Working Application- **GitHub (code + lightweight demo subset):** https://github.com/ucerd/Cricket_Analyzer - **Live Streamlit App:** http://cloud.pakhpc.com:8502/ > Note: The GitHub repository contains the complete codebase and a **functional subset (~1.16 GB)** for quick testing. > This Zenodo deposit contains the **full dataset (~20 GB)** used for complete training and evaluation. --- ## What this dataset containsThis Zenodo archive includes:- **Raw cricket delivery videos and/or curated extracted frames** used in training and evaluation (as per the manuscript)- **YOLO-format annotations** for required classes (e.g., *ball* and *wicket*)- **Metadata** describing capture conditions (venue type, pitch surface, lighting/weather, camera type)- **Train/validation/test split definition** (either via split lists and/or directory-based partitioning)- *(Recommended for PLOS ONE)* Any **CSV/JSON tables** containing the underlying values behind figures/tables reported in the manuscript --- ## Dataset organization (high-level)The dataset is structured to preserve real-world variability. Typical hierarchy includes:- ball type/color (e.g., `red_balls`)- venue (e.g., `box_cricket`)- pitch surface (e.g., `cemented_pitch`)- lighting/weather (e.g., `cloudy_weather`, `dark_weather`)- camera type (e.g., `gopro_camera`, `iphone_camera`) Images/frames and labels mirror each other in standard YOLO layout:- `images/train/...`, `images/val/...` (and `images/test/...` if applicable)- `labels/train/...`, `labels/val/...` (and `labels/test/...` if applicable) --- ## How to reproduce results (training + inference)1. **Download and extract** this Zenodo dataset locally.2. In the GitHub repo, update your `data.yaml` to point to the extracted dataset path.3. Run training/fine-tuning using the scripts provided in the repository.4. Run inference: - either via the **Streamlit app** (recommended), - or via the provided inference script to compute detection + trajectory + speed. GitHub repository (code + usage): https://github.com/ucerd/Cricket_Analyzer --- ## Ethics & privacyThe dataset is released for research and reproducibility. The recordings were captured with appropriate permission for research use. No intentionally identifying personal metadata is included in the released files. If any file is found to contain sensitive/identifying content, please contact the maintainer for remediation. --- ## Versioning & integrity- **Dataset version:** v1.0.0 (update if needed)- This Zenodo record should be treated as the **version of record** for the dataset used in the manuscript.- *(Recommended)* A `CHECKSUMS.txt` (SHA256) manifest can be included for file integrity verification. --- ## License- **Recommended dataset license:** CC BY 4.0 (Creative Commons Attribution 4.0) --- ## CitationPlease cite this dataset as:**Cricket_Analyzer Full Dataset (v1.0.0), Zenodo, DOI: [Zenodo DOI]** --- ## Maintainer / Corresponding Author**Prof. Dr. Tassadaq Hussain** Email: tassadaq@ucerd.com