1) Here are the code used in the following paper. Sample images are shown in the figure below.

S. Fan, M. Jiang, B. Koenig, Z. Shen, M. Kankanhali, Q.Zhao. "The Role of Visual Attention in Sentiment Prediction", ACM Multimedia, Mountain View, CA, USA, 2017.

We share the first test of pre-trained CNN (VGG-19) using Caffe for sentiment prediction. You could download the model here. The architecture of the proposed CNN is shown in the image below.

2) Here are the code and model used in the following paper.

S. Fan, Z. Shen, B. Koenig, T. Ng, M. Kankanhali, When and why static images are more effective than videos. Submitted manuscript.

We share the code of CNN (C3D) using Keras. You could download the code here.