This study presents multidimensional deep learning approaches on hyperspectral images. Storing, processing and interpreting hyperspectral data is very difficult due to its complexity and processing load. Consequently, conventional classifiers are not feasible to extract distinctive features. In order to present efficient classifiers, we utilize deep learning and present Convolutional Neural Network (CNN) approaches. In this study, we evaluate one-dimensional, two-dimensional and three-dimensional convolution model approaches that can present efficient classification performance. Within the scope of the study, samples of widely used hyperspectral data sets are classified by using one-dimensional, two-dimensional and three-dimensional convolutional neural networks by extracting spatial, spectral and spatial-spectral features. All the features provided by hyperspectral sensors are included in the classification by using both separate and joint spectral and spatial features. As a result, our studies have shown that our three-dimensional Convolutional Neural Networks have achieved higher classification rates compared to the state of art models. The accuracy performance of our three dimensional convolutional neural network is able to converge to 100% during classification.