Lighting is one of the largest uses of energy in buildings in the United States and around the world. A building lighting audit is one step to reducing the total energy consumption of a facility. This report presents a fast and automated method to detect, classify, and measure lighting in buildings without installation of any monitoring devices and without a building floor plan. With this system, the manual, tedious, and error-prone process of counting and marking all lights on a pre-given floor plan is replaced by two quick walkthroughs of the area with a custom collection device. The device consists of a ceiling-facing Canon DSLR camera, a Google Project Tango tablet, and an Ocean Optics Spark-VIS spectrometer, all mounted together in an easy to carry package. Using the data collected with this device, lights are detected through a series of image processing operations on pictures acquired by the camera and then tracked through multiple frames. Using the precise position and pose information from the Tango tablet and correspondences between sequential images, the dimensions of each light are estimated. Finally, using a neural network, the lights are classified based on their emission spectra. Operation was tested with four data sets recorded in Cory Hall and the Valley Life Sciences Building on the University of California, Berkeley campus. This system detected lights with a 7% error rate, classified them with a 14% error rate, and estimated surface area within a factor of two.