This paper introduces an EEG (Electroencephalography) dataset collected using the Rapid Serial Visual Presentation (RSVP) paradigm combined with an HMD (head-mounted display), to simulate real-world conditions for Brain-Computer Interfaces (BCIs). EEG signals were recorded under both clean and intentionally contaminated conditions, where participants engaged in activities to intentionally introduce artifacts deleterious to system performance such as head movements, body movements, and talking, simulating real-world sources of noise. Unlike the conventional RSVP-BCI method that displays stimuli on a monitor, we investigate the use of a head-mounted display, and compare both approaches under clean and noisy conditions to assess their performance differences. Target versus non-target classification using multiple machine learning models revealed a decline in ROC-AUC scores under noisy conditions compared to clean settings. Additionally, our comparative analysis between HMD-based and monitor-based BCIs, supported by paired t-tests, confirmed that the differences in classification performance were not statistically significant, reinforcing the feasibility of using HMDs for real-world BCI applications. By incorporating a head-mounted display setup, this dataset provides a valuable resource for developing artifact-resistant Brain-Computer Interfaces applicable to gaming, metaverse environments, and assistive technologies. Researchers can leverage this dataset to explore signal-denoising techniques and enhance the robustness of EEG-based BCIs in naturalistic conditions that use HMDs. The datasets presented in this study can be found at: \url{https://zenodo.org/records/14750342}