: 43. IR Surface Reflectance Estimation and Material Type Recognition Using Two-Stream Net and Kinect Camera
Add to Agenda
Session TimeSunday, 28 July 20191:30pm - 5:30pmMonday, 29 July 201912:15pm - 1:15pmWednesday, 31 July 201912:15pm - 1:15pm
LocationSouth Hall J
DescriptionPOSTER SESSIONS: MONDAY, 29 JULY and WEDNESDAY, 31 JULY, 12:15-1:15 PM
Material-type recognition using color or light field camera has been studied and collect diverse visual characteristics of each material type. However, visual pattern-based approaches for material-type recognition without direct acquisition of surface reflectance show limited performance. We propose IR surface reflectance estimation using an off-the-shelf ToF (Time-of-flight) active sensor, such as Kinect, and perform surface material-type recognition based on both color and reflectance clues. Two stream deep neural networks consist of convolutional neural network encoding visual clues, and recurrent neural network encoding reflectance characteristic is proposed for material classification.