Water electrolysis is one of the key processes for carbon-free hydrogen (H-2) generation, yet it inherently presents the risk of oxygen (O-2) permeation, a significant concern that must be addressed. Contrary to external atmospheric conditions, the environment within a water electrolysis system is characterized exclusively by high-purity H-2, little O-2, and moisture (H2O). However, there has been limited research on sensors capable of detecting O-2 crossover in such a pure H-2 atmosphere under high humidity conditions. In this study, we developed an electrolysis O-2 monitoring (E-OM) sensor based on a semiconducting metal oxide (SMO) that can operate at a room-temperature with such high humidity through photoactivation with 365 nm LED light. A porous In2O3 nanofilm, prepared by glancing angle deposition (GLAD), was used as the gas sensing material, and copper nanoparticles (Cu NPs) were decorated on the In2O3 surface through electron beam evaporation to accelerate the O-2 adsorption. Remarkably, the E-OM sensor showed a humidity-independent response to O-2 gas in the H-2 atmosphere. Meanwhile, since the response of the E-OM sensor itself was not fast enough for water electrolysis system application, convolutional neural network (CNN)-based signal processing in the time domain was applied. As a result, our E-OM sensor system could predict O-2 concentrations from 0 to 2 vol% in an H-2 environment with a relative humidity (RH) of 30-90 %, and the detection time was under 5 s. This development could enable safe, rapid, and cost-effective monitoring of O-2 crossover in H-2 production infrastructure.