TensorFlow and OpenCV are open source libraries used for machine learning and image manipulation, respectively. Using these libraries, we develop a Python-based program that utilizes trained convolutional neural networks for facial recognition. As needs for object recognition increases in various applications such as autonomous vehicles, medical imaging, and facial recognition, it is useful to know what effect color spaces may have on image recognition models. The purpose of our project is to run basic tests with simple facial recognition models to verify which model performs the best when inputs are subject to uniform image manipulation. Three models with identical network architectures are trained with RGB, HSV, and LAB color space images. Images are then uniformly darkened and brightened in the RGB color space before being converted to the separate color spaces and run in the models trained earlier. We found that the RGB color space model responds the best to image manipulations of this kind using our dataset. The HSV color space model had the second-best performance and the LAB color space model suffered significantly. All three models performed worse when images were darkened as opposed to brightened. Answers are proposed as to why some color space models perform poorly. The paper will also look at performance metrics while training the neural network for different runtime environments when using Google Colab.