CrickiWiki

Cricket Video-to-Text Summarization Using Neural Networks Major Project

This project takes the cricket video as input and generates a concise and contextual summary for the cricket match video utilizing only visual information.

RESULTS:

ABSTRACT

This project introduces an inventive application of neural networks in cricket sports video-to-text summarization, focusing on the extraction and representation of key highlights and events. The proposed system aims to transform the traditional approach to summarizing cricket match videos by harnessing the power of neural network architectures, including VGG-16 Convolutional Neural Networks (CNNs), Optical Character Recognition (OCR) and Long Short-Term Memory Recurrent Neural Networks (RNNs). By converting visual cues into concise textual summaries, the system offers an efficient solution for summarization, enhancing accessibility for a diverse audience. This innovative approach not only caters to cricket enthusiasts but also presents an invaluable tool for coaches, analysts, and sports professionals. The system's ability to distill complex match details into easily digestible textual summaries offer a streamlined approach to glean insights, strategize, and review match performances effectively. The integration of neural networks adds a layer of sophistication to the summarization process, offering a unique perspective on information extraction from cricket sports videos. Ultimately, this project aims to contribute to a new era of accessible and streamlined content consumption, free from the complexities of in-depth sports.

Submitted By

• Name: Team-02

• Program: B. Tech CSE - AIML

• Institution: MLR Institute of Technology

  Team Members

Here are the details of our team members:

Member Name	Email	GitHub
Dhatrika Kamal Kumar	[email protected]	@Dhatrika-Kamal-Kumar
Maddhe Sai Prashanth	[email protected]	@Maddhe-Sai-Prashanth
Vuppala Praneeth Kumar	[email protected]	@Vuppala-Praneeth-Kumar
I Nikhil Sri Sai Teja	[email protected]	@I-Nikhil-Sri-Sai-Teja

SOFTWARE REQUIREMENTS:

1. Operating System: Compatible with Windows 10, macOS Mojave (10.14) or later, or popular Linux distributions such as Ubuntu 18.04 LTS or newer versions.
2. Python Environment: Python 3.x is installed with essential libraries such as TensorFlow, Keras, scikit-learn, NumPy, NLTK, and OpenCV for machine learning, natural language processing, and image processing tasks.
3. Development Tools: Integrated Development Environments (IDEs) such as PyCharm, Jupyter Notebook, or VSCode for coding, debugging, and experimentation.
4. Version Control: Git installed for version control management, facilitating collaboration and tracking changes in code and project files.
5. External Libraries and Models: Installation of additional libraries and models such as Paddle OCR, BART (Bidirectional and Auto-Regressive Transformers), and pretrained models like VGG16 for image processing and text summarization tasks.
6. Internet Connectivity: High-speed internet connection for accessing online resources, downloading additional datasets, and cricket match footage.

HARDWARE REQUIREMENTS

1. CPU: A modern multi-core processor (Intel Core i5 or equivalent) to handle computational tasks efficiently.
2. GPU: A dedicated graphics processing unit (NVIDIA GeForce GTX 1060 or equivalent) with CUDA support for accelerated deep learning computations, especially for training large neural network models.
3. RAM: A minimum of 8GB of RAM (16GB recommended) to ensure smooth processing of large datasets and model training operations.
4. Storage: Adequate storage space (at least 500GB HDD or SSD) for storing video datasets, image frames, trained models, and intermediate data files.

Libraries Used:

torch==1.10.0
ultralytics==8.0.3
opencv-python==4.5.5.64
numpy==1.21.4
pillow==8.4.0
pandas==1.3.4
paddlepaddle==2.2.2
paddleocr==2.0.2
matplotlib==3.5.0
transformers==4.12.3
nltk==3.6.3
sentencepiece==0.1.96

How to Run Locally

To run this project on your local machine, follow these steps:

Clone the project repository:

git clone [(https://github.com/CSM02-Cricket-Video-to-Text-Summarization-Major.git)]

Go to the project directory

cd /CSM02-Cricket-Video-to-Text-Summarization-Major

Install Requirements

pip3 install -r requirements.txt

System Requirements:

- Python (3.10 recommended)
- OpenCV
- YOLOV8
- VGG16
- LSTM
- PaddleOCR
- Distilbart
- torch
- ultralytics
- numpy
- pillow
- pandas
- matplotlib
- transformers
- nltk
- sentencepiece

HOW TO IMPLEMENT THIS PROJECT:

allmodules.py file consists of all the code to generate the summary of input cricket video

1. Run app.py (flask program)
2. It renders index.html
3. Click the Upload button to upload a cricket video as input
4. Then click the Submit button to process the summary for the input video
5. After the processing, click the Generate button to get the textual summary of the uploaded video

Video Demonstration

For a detailed walkthrough of our project, watch the video here. Model Files: here

Contributing

Pull requests are welcome. For major changes, please open an issue first to discuss what would you like to change.