COMPUTER-IMPLEMENTED METHOD FOR 3D LOCALIZATION OF AN OBJECT BASED ON IMAGE DATA AND DEPTH DATA

摘要

The invention relates to a computer-implemented method for 3D localization of an object based on image data and depth data indicating distance information for pixels of an image represented by the image data, wherein the method applies a convolutional neural network (1) with a first set (2) of consecutive layers (21), a second set (3) of consecutive layers (31) and a third set (5) of consecutive layers (51), each layer being configured with one or more filters, wherein the convolutional neural network (1) is trained to associate an identification/ classification of an object and corresponding 3D localization data for the identified/classified object to an image data item and a depth data item, comprising the steps of:- Extracting one or more image data features from the image data applying the first set (2) of layers (21) of the convolutional neural network (1);- Extracting one or more depth data features from the depth data applying the second set (3) of layers (31) of the convolutional neural network (1);- Fusing the one or more image data features and the one or more depth data features to obtain at least one fused feature map;- Processing the at least one fused feature map by applying the third set (5) of layers (51) of the convolutional neural network (1) to identify/classify the object and to provide the 3D localization data for the identified/classified object, wherein the 3D localization data includes object reference point data.