In order for a system to learn a model for object recognition, it must have a lot of positive images to learn from. Because of this, datasets of similar objects are built to train the model. These object datasets used for learning models are best when large, diverse and have annotations. But the process of obtaining the images and creating the annotations often times take a long time, and are costly. We use a method that obtains many images of the same objects in different angles very quickly and then reconstructs those images into a 3D model. We then use the 3D reconstruction of these images of an object to connect information about the different images of the same object together. We use that information to annotate all of the images taken very quickly and cheaply. These annotated images are then used to train the model.

This thesis presents and evaluates different methods to semantically segment 3D-models by rendered 2D-views. The 2D-views are segmented separately and then merged together. The thesis evaluates three different merge strategies, two different classification architectures, how many views should be rendered and how these rendered views should be arranged. The results are evaluated both quantitatively and qualitatively and then compared with the current classifier at Vricon presented in [30]. The conclusion of this thesis is that there is a performance gain to be had using this method. The best model was using two views and attains an accuracy of 90.89% which can be compared with 84.52% achieved by the single view network from [30]. The best nine view system achieved a 87.72%. The difference in accuracy between the two and the nine view system is attributed to the higher quality mesh on the sunny side of objects, which typically is the south side. The thesis provides a proof of concept and there are still many areas where the system can be improved. One of them being the extraction of training data which seemingly would have a huge impact on the performance.