First-person activity recognition: how to generalize to unseen users?
Tipus de documentProjecte Final de Màster Oficial
Condicions d'accésAccés obert
Recent advances in wearable technology, accompanied by the decreasing cost of data storage and increase of data availability have made possible to take pictures everywhere at every time. Wearable cameras are nowadays among the most popular wearable devices. Besides leisure, wearable cameras are attracting a lot of attention for the improvement of working conditions, productivity and safety monitoring. Since the collected data can be potentially used for memory training and extracting lifestyle patterns useful to prevent noncommunicable diseases as obesity, they are being investigated in the context of Preventive Medicine. Most of these applications require to automatically recognize the ability performed by the user. This work aims to make a step forwards towards activity recognition from photo-streams captured by a wearable camera by developing a method that allows to label new images with minial effort from the user and generalize well for unseen users.
En col·laboració amb la Universitat de Barcelona (UB) i la Universitat Rovira i Virgili (URV)