Author(s):
1. Hassham Jamil:
Department of Computer Science, University of Engineering and Technology, Taxila, Pakistan
2. Fatima Khalid:
Faculty of Computer Science University of Engineering and Technology, Taxila, Pakistan
3. Fatima Khalid:
Faculty of Computer Science and Engineering, GIK Institute of Engineering Sciences and Technology, Topi, Pakistan
Abstract:
Cricket shot classification in sports video analysis is challenging due to factors like camera angles, quick scene changes, and varying lighting conditions. Analyzing cricket videos can be tricky because it's hard to capture the fine details of different batting techniques. This research tackles this challenge by proposing a new approach specifically designed for cricket. Imagine two models, like teammates working together. One model focuses on specific details (Model S), while the other looks at the bigger picture (Model R). This research compared these models using different approaches (activation functions) to see which worked best for recognizing various cricket shots. In addition to these two models, a "fused" model was created. Think of it as these two models joining forces, with extra layers added to improve their ability to analyze the videos. This involved extracting key moments (frames) from the videos, training the models using a large dataset of cricket footage, and then combining them for even better results. This new approach achieved an amazing accuracy of 96.3% in classifying different cricket shots. This wasn't just a one-off test either. Additional checks confirmed it seems to work well in sports video analysis, especially in the fast-paced world of cricket. Overall, this research demonstrates the power of a specialized model (Fused Truncated ConvNeXT) for accurately figuring cricket shots. It opens doors for further advancements in video analysis, this lets us see the skill and different ways batsmen hit the ball in cricket.
Page(s):
1-1
DOI:
DOI not available
Published:
Journal: Second International Conference on Computing Technologies, Tools and Applications (ICTAPP-24), June 4-6,2024 (Abstract Book), Volume: 0, Issue: 0, Year: 2024
Keywords:
deep learning
,
Feature Fusion
,
MTCNN
,
ConvNeXT Tiny
,
Cricket Video Analysis
,
Shot Classification
,
Sports Video Analysis
References:
References are not available for this document.
Citations
Citations are not available for this document.