Indoor positioning in aircraft cabins presents fundamental challenges arising from severe multipath propagation, non-line-of-sight conditions, and metallic fuselage geometry that degrade radio-based positioning methods. This study validates a residual neural network (ResNet) based deep learning approach for aircraft cabin localization through real-world measurements in an A320 cabin mockup. The methodology employs dual-technology ranging measurements from Ultra-Wideband and Bluetooth Low Energy, transforming range observations into spatial likelihood representations processed by a ResNet. Experimental validation encompasses 19 distributed measurement positions, evaluated against three baseline methods: iterative least squares, robust least squares with Huber loss, and Bayesian grid filtering. ResNet achieved an overall median positioning error of 0.177 m, achieving lower positioning errors than all three baseline methods. Results confirm that likelihood-based neural network positioning is viable for operational aircraft cabin deployment while identifying performance dependencies on anchor visibility, measurement height, and propagation conditions. The original data is openly available.