<p>Resource estimation of a mining project is an important step in mine planning process. Predicting mineral grades with precision is a crucial aspect which significantly influences the economic assessment of mining projects. Existing approaches work on the basis of geometry and geostatistics. In case of geological complexity in a deposit, lot of expertise and experiences are required to predict the grade using geostatistical techniques. In this psaper, we present Artificial Neural Network (ANN) model for ore grade estimation. The models are trained with one and two hidden layers using Relu and sigmoid activation functions with different number of neurons. The models are trained with the combination of RMSprop and Stochastic Gradient Descent (SGD) algorithms. The models were trained by using sample coordinates (X, Y, and Z) and grade (%Fe) obtained from the drill hole data. A case study in Indian iron ore deposit shows that using sigmoid activation function with two hidden layers, the proposed ANN model can predict better results with R² of 0.71 and MSE of 3.60. The suggested ANN model is adaptable for use with other datasets containing similar types of deposits. </p>