<![CDATA[Advancements in digital technology have transformed conventional attendance systems into more secure and automated solutions. However, manual and online-based systems remain vulnerable to fraud, such as proxy attendance and false records. This study designs a digital attendance system using face recognition technology based on the Convolutional Neural Network method. The process begins by capturing facial images via camera, followed by preprocessing steps including grayscale conversion, face detection using Haar Cascade, and resizing images to 100x100 pixels. The CNN model is trained with the preprocessed dataset and saved in .joblib format for real-time face identification. Attendance is automatically recorded in a CSV file. Testing was conducted based on dataset size, distance, and face position relative to the camera. Results show that accuracy improves with more training data. Using 200 images per individual yielded the best balance of accuracy, speed, and storage efficiency unlike 50 images, which often failed, or 500 images, which required long training times and large storage. Lighting quality also significantly impacts recognition accuracy, poor or uneven lighting leads to unclear facial features. Thus, proper lighting is essential. This study demonstrates that CNN effectively supports the digital transformation of attendance systems, making them more accurate, efficient, and fraud-resistant.]]>
Copyrights © 2025
