The trend toward the use of light-emitting diodes (LED) for photometric and colorimetric measurements has put pressure on studies that aim to improve spectral uncertainties with calibration methodologies. Recent studies have shown good results using artificial neural networks (ANN) to calibrate a colorimeter; however, this methodology presents the difficulty of finding the segmentation of the input variables for ANN training. We propose the use of a new calibration methodology for a photometer, an ANN that transforms the output of the sensor into luminous transmittance without the need of a filter in a prototype that uses an RGB sensor and a white LED as an illuminant to verify the possibility of calculating the luminous transmittance of a sample. A 14,031 spectra dataset was built that covered the entire input range for training the final ANN, a four-input multilayer perceptron, and 20 neurons in the hidden layer. The ANN was validated with errors smaller than a class L photometer. The prototype was tested for measuring luminous transmittance, under the D₆₅ illuminant of colored filters samples with all results within the ±3  %   points error range. The prototype overcame commercial meters with better results. The ANN can effectively calculate luminous transmittance and can be tested to calculate other photometric values that use different weighting functions with the same hardware. The methodology presents an option for a meter with a correction method that depends only on software. This allows the development of a compact and low-cost photometer.