This project was developed as the final submission for the "Belajar Machine Learning untuk Pemula" (Learning Machine Learning for Beginners) course at Dicoding. The primary objective was to build a robust image classification model capable of accurately recognizing hand gestures representing rock, paper, or scissors. The development process involved a complete machine learning pipeline. It started with dataset preparation, including an optimal train-validation split. To ensure the model's ability to generalize and prevent overfitting, extensive Image Augmentation techniques were implemented using Image Data Generators. The core of the project is a custom Sequential model architecture that was trained to achieve an accuracy of over 85%. Furthermore, the project includes an interactive prediction feature within Google Colab, allowing users to upload custom images to be classified by the trained model in real-time.