Our paper on "Emotions in Context" accepted at CVPR 2017

We present a database for studying how to model context to understand people emotional states. We show promising results at estimating 26 affective categories and continuous dimensions (pdf, Project Page)

Our work on "Class Activation Map" accepted at CVPR 2016

We revisit the global average pooling layer and shed light on how it explicitly enables the convolutional neural network to have remarkable localization ability despite being trained on image-level labels (Project Page).

Scene recognition demo

Given a picture our system predicts the scene category and some other attibutes. It also provides a heatmap that indicates the region of the image that supports the ouputs.

Understanding the representations learned by CNNs

We found that object detectors emerged in a CNN trained for scene recognition. For more information check our paper: B. Zhou, A. Khosla, A. Lapedriza, A. Oliva, and A. Torralba. “Object Detectors Emerge in Deep Scene CNNs.” International Conference on Learning Representations (ICLR) oral, 2015. (PDF).

Project page of Places Database

You can download the database and the pretrained network PlacesCNN. More details can be found in our paper: B. Zhou, A. Lapedriza, J. Xiao, A. Torralba, and A. Oliva. “Learning Deep Features for Scene Recognition using Places Database.” Advances in Neural Information Processing Systems 27 (NIPS), 2014. (PDF).