In this study, visible RGB-based vegetation indices (VIs) from UAV high spatial resolution (1.9 cm) remote sensing images were used for modeling shoot biomass of two Brazilian wheat varieties (TBIO Toruk and BRS Parrudo). The approach consists of a combination of Artificial Neural Network (ANN) with several Vegetation Indices to model the measured crop biomass at different growth stages. Several vegetation indices were implemented: NGRDI (Normalized Green-Red Difference Index), CIVE (Color Index of Vegetation Extraction), ExG (Excess green) SCOM (Simplified Combined Index) and a new index that we called ExRM (Excess red modified). An experiment containing 120 test plots was designed to assess wheat growth and test the vegetation indices (VIs) performance by correlating them with measured shoot dry biomass. Variability in crop growth was created for all test areas by varying nitrogen availability. For determining shoot biomass, plants were sampled at two different crop growth stages: V6 (stage of six fully developed leaves) and flowering. These measures were considered as the golden standard for the biomass model estimators. The images of the test areas were captured using an UAV flying at 50 meters above ground, a mosaic was created, and then the regions of interest were segmented. An ANN was trained to predict Biomass using Vegetation Indices as the features. We also compared the results with a linear regression to estimate shoot biomass from the VIs. The accuracy of the estimated model was evaluated based on the coefficient of determination (R2). The best result for the cultivar BRS Parrudo was R2=0.81 obtained using ANN and all VI as features versus biomass. For the cultivar TBIO Toruk the best result was R2=0.86, modeling the biomass with four selected VIs. Our research results indicate that the proposed estimation model, based on RGB images and ANN, can be used in precision agriculture for predicting the spatial variability of shoot biomass, considering the two wheat cultivars tested.