Cross-Model Interpretation of Machine Learning-Based Automatic Baybayin Script Recognition Systems

In the past years different image recognition models have been used and while there are models that are more dominant in all image recognition task, remaining models are still as important, and some perform even better than the current generation models. This study compares four models with the assigned task to recognize Baybayin script characters. Results show that CNN outperformed the other models with a result of 89% accuracy compared to the new Vision transformer that has 52% accuracy. After some investigation the researchers found that transformer-based models for computer vision demands a high requirement in term of processing power and requires a large amount of data in order to reach its peak performance.