Deep learning has become widely used in image analysis. Transfer learning can make use of information from other data sets for the analysis of this data set. When there is a small number of images at hand, the deep learning method will conduct transfer learning, which means using trained models or coefficients from other data sets. This is in contrast to deep learning with most model parameters re-estimated. Transfer learning will make use of trained models from other data sets and then apply them to images of this dataset to extract high-level features. High-level features can be fed into traditional machine learning models including a Neural network. We compare a range of combinations of traditional machine learning with deep learning for multi-view image analysis, with the objective of improving image analysis performances. The proposed methods have been applied to multi-view plant phenotyping data to evaluate the performance of various methods.