Hematoxylin-eosin (H&E) stained tissue slides are routinely obtained from each patient in the pathology department. These slides are also called histopathology images. Examination of histopathological images is the gold standard in the diagnosis of tissues. However, manual analysis of histopathology images is a serious and time-consuming task for pathologists. Therefore, it is essential to automatically develop medical models to classify tissues or structures of interest in pathological images. For this purpose, studies were conducted on classifying nine different tissue types from histopathology images. This study used opensource image sets obtained from 86 patients as training set and 50 patients as test set. The comparison was made on the results of the two different approaches by training convolutional neural networks and machine learning models for the automatic classification of tissues. At the same time, the training of the models was done for two different classification tasks. The first is binary classification as normal and tumor tissue classes and the other is multi-classification as nine different tissue types: adipose, background, debris, stroma, lymphocytes, mucus, smooth muscle, normal colon mucosa, and tumor. Accuracy results up to 0.93 and 0.90 were obtained for binary and multiple classifications, respectively, using convolutional neural networks from the test sets. By extracting the histogram features of the directed gradients and using the Random Forest classifier, accuracy rates up to 0.75 and 0.44 were obtained for binary and multiple classifications, respectively. The results demonstrate the generalizability of our convolutional neural networks model for both binary and multi-tissue classification and the extremely promising potential of embedding them into a practical histopathological diagnosis workflow system.