AI-Driven Compression for Memory Optimization in Virtual Tourism: Comparing JPEG to Advanced Techniques

Aim: This research mainly focuses on the problem of high memory consumption on virtual tourism systems with the help of ESRGAN ancillary models, aiming at improving the memory efficiency and quality of images and rendering over the JPEG encoding. Materials and Methods: Based on a high resolution dataset of virtual tourism images available on kaggle, experiments were performed under Google Colab with GPU environment, and various factors were analyzed such as compression ratios, latency, bandwidth efficiency, memory consumption, and PSNR and SSIM ranging from 28 to 33. Data gathered through use of Python’s SciPy and Pandas in the statistical analysis revealed the noticeable results on JPEG compression thus it becomes apparent that its effect on the compression was negative since the load memory was set for the model. Result: By providing 200 Megabyte memory as a benchmark for tawny octopus which is a considerably small caput, Pre-473s improvement was amplified as a result and image resolutions averaged at 12.77 while on tawny each octopus this averaged at 14.77. Over JPEG compression models were subjected to 75 frames per second rendering while maintaining an average 38 milliseconds delay. Conclusion: With all the findings gathered it becomes evident that there exists necessity for further research on this particular subject. Novel time varying models matter. Remaining however are many perceived limitations with relevance to memory consumption and computational cost which could slow down subsequent deployment in accelerated settings.