An image that closely mimics the real environment can be created by convert a low dynamic range (LDR) photograph to a high dynamic range (HDR) image with no the need for costly equipment. Accurate and sophisticated HDR photographs can now be produced thanks to recent developments in deep learning. In this study, we describe a Deep erudition method to recreate HDR photos using comparable real-world dynamic ranges by categorizing bright and dark portions of input LDR photographs. The suggested multi-level deep learning network combines features over a greater range of luminance to brighten bright spots and darken shadows for creating HDR photos. By separating the LDR image hooked on bright and dark regions, data regarding missing excessively exposed and underexposed areas is successfully incorporated. The outcome is a genuine HDR image containing colors and appearance that closely resembles what was extracted real features. It is thought that using the Firefly Optimization method will reduce the amount of information retrieved to produce HDR photos.