This paper introduces PIXIE (Penalised Pixel Intersection Error), a novel loss function aimed at improving model performance through the penalization of error pixels and alignment of the predicted mask with the ground truth mask. We explore the effectiveness of PIXIE alongside U-Net and DeepLabV3 networks. In addition to novel loss functions, this paper conducts model selection studies to understand overestimations and underestimations. We conduct a comprehensive evaluation using diverse datasets spanning medical imaging (breast tumor, COVID-19, Brain MRI) and remote sensing domains (forest fire, water bodies satellite images). Performance metrics, such as Intersection over Union (IoU), Dice Coefficient (DC), Area Error Ratio (AER), precision, and recall are quantitatively assessed. Moreover, we compare PIXIE with traditional loss functions like Jacard loss, Focal loss, and Binary crossentropy. PIXIE demonstrates comparable performance to the traditional loss functions in certain metrics and outperforms them in others, establishing itself as a leading approach in achieving exceptional results across all examined segmentation datasets. These findings represent a substantial contribution to the practical evolution of semantic segmentation in computer vision, offering essential insights into the optimization of loss functions for the development of accurate and robust models.

This paper introduces a novel approach to a complete contactless biometric system that takes a finger photo image as an input and performs various image processing techniques and authenticates the fingerprints for an easy, non-invasive system that is efficient and robust. While contact-based fingerprint recognition systems have produced ground-breaking achievements, these systems face issues with latent fingerprints, sensor degradation brought on by frequent physical touch, and hygiene issues. Thus, the next step towards solving these issues is developing a contactless system that counters all the mentioned issues as faced by a contact-based fingerprint recognition system. This paper introduces a novel deep learning architecture that fuses the scattering wavelet transform making it lightweight and computationally efficient. A unique combination of the Siamese network integrated with wavelets and the traditional minutiae-based approach builds the core framework for the recognition system. The combination of these techniques allows the system to perform fingerprint recognition with high accuracy. This approach performs with an Equal Error Rate (EER) of 2.5% on the IITI-CFD, 2.5% on the PolyU 2D Contactless Dataset, and 3.76% on the IITB Touchless Fingerprint Dataset. The system aims to achieve a balance between efficiency and economy.