Unlocking the Future: Palm Recognition with Convolutional Neural Networks(CNN)

• Published in: Journal of neonatal surgery Vol. 14; no. 15S; pp. 1447 - 1457
• Main Authors: Reddy, Viswanathan Ramasamy, Singh, Sukham Romen, Guruva Reddy, Elangovan, E. Punarselvam, E. Punarselvam, T. Vengatesh, T. Vengatesh
• Format: Journal Article
• Language: English
• Published: 16.04.2025
• ISSN: 2226-0439; 2226-0439
• DOI: 10.63682/jns.v14i15S.3863

Palm recognition, a subset of biometrics, has received a lot of attention due of its potential utility in many different domains, including security systems and human-computer interface. In this study, convolutional neural networks (CNNs) are being utilised to investigate the development of a palm detection system. Deep learning will be used to detect and categorise palm orientation in digital photographs. Data collection, which entails gathering a broad dataset of precisely labelled palm pictures for supervised learning, is an important component of the project. We manage data preprocessing and prepare datasets for model training by employing techniques such as picture scaling, pixel normalisation, and data augmentation. Our technology is built around a CNN model architecture, which generates a neural network capable of automatically collecting information from palm images and conducting smart classification. This design employs convolutional layers for feature extraction and fully linked layers for classification. During the training phase, we investigate the technique, optimizer selection, loss function selection, and hyperparameter optimisation. We pay close attention to two things: monitoring the model's performance on the validation set and putting countermeasures in place to avoid overfitting. The evaluation section provides information about the model's precision and generalizability. We review the results of testing on the validation and test datasets while keeping the problems and limits in mind. Our efforts to fine-tune the model involve adjusting hyperparameters and researching data augmentation approaches, all with the goal of improving model performance. During the inference phase, the trained model's potential in real-world situations is highlighted, demonstrating how it might be applied in practise. Our palm recognition technique paves the path for future biometric authentication use, with potential applications in security, access management, and human-computer interface.