Deep learning has become a popular topic in machine learning, with the use of deep neural networks producing state-of-the-art results in many domains ranging from machines generating their own cryptographic techniques for keeping messages secret to beating the best human players at the game of Go.
In our research group, we have also been working on deep learning techniques, applied in the domain of computer vision. This has allowed us to discover patterns and events in the physical world by analysis of multiple streams of sensor data.
This can provide beneﬁt to society in more than just surveillance applications by focusing on automated means for social scientists, anthropologists and marketing experts to detect macroscopic trends and changes in the general population. Our goal was to complement analogous eﬀorts in documenting trends in the digital world, such as those in social media monitoring with automated, deep learning methods.
Using deep learning techniques, we have shown how the contents of a street webcam can provide information about patterns in clothing and their relation to weather information. In particular, we analysed a large time series of street webcam images, using a deep network trained for garment detection, and demonstrated how the distribution of clothing types over time signiﬁcantly correlates to weather and temporal patterns.
Furthermore, we have used deep learning techniques to train a classifier for detecting the gender of faces in images, achieving record-breaking accuracy of 98.9% on the Labeled Faces in the Wild (LFW) data set and 91.34% on the Images of Groups (GROUPS) data set in a cross-database setting.
Future work on deep learning will also investigate the application of deep neural networks for natural language processing and their application in text analysis.