Neural Radiance Fields (NeRF) is a recent method that has drawn a lot of attention over the 3D scene reconstruction field. Once trained on a set of 2D images taken from a complex scene, it can render novel views. It has been emerging as an alternative to the classical 3D reconstruction methods. Even though the NeRF is successful in general, there is no clear information on whether NeRF is also successful in limited settings. This thesis is aimed at conducting a systematic evaluation on the limitations of the state of the art NeRF based 3D scene reconstruction methods with varying low resolution and different color mode input images. The reasonings behind these experiments are to get an insight on whether
current models can be used with the real low resolution images that are captured from the real world systems such as small quadrotors and what is the least amount of information that is necessary in order to be able to use NeRF based methods successfully. To be able to answer these questions, a test environment is created to test Instant NGP and NeRF together with input images with varying color modes and low resolutions gathered from the Habitat Lab simulation and poses estimated by COLMAP with these input imagesets. Output of the experiments are reported in PSNR, SSIM and LPIPS metrics. Results have shown that (i) higher resolution images create reconstructions with better quality, (ii) color mode affects each metric differently, and (iii) the intrinsics that are taken from COLMAP using the high resolution images also improves the quality of reconstruction for low resolution images.