Humans can understand the 3D environment in a very efficient way by only focusing on the important parts. Meanwhile, directly processing the 3D representation of the world can be computationally expensive. A better representation of our living world can help automated systems to understand the world with higher certainty. Compared to 2D images, 3D images give us a better representation of this world. We are living in a three-dimensional world.
Finally, the use of segmentation as part of our pipeline increases detection accuracy, while providing at the same time 3D instance segmentation. This is important since our methods can also work well in low lighting conditions, or with sensors that do not acquire RGB images. At the same time, we can provide segmentation and detection results using depth only images, with accuracy comparable to RGB-D-based systems. Results on the SUN RGB-D dataset show that our RGB-D-based system’s 3D inference is much faster than state-of-the-art methods, without a significant loss of accuracy. This system generates frustums from 2D detection results, proposes 3D candidate voxelized images for each frustum, and uses a 3D convolutional neural network (CNN) based on these candidates voxelized images to perform the 3D instance segmentation and object detection. A 3D convolutional-based system, named Frustum VoxNet, is proposed. First, we detect 2D objects based on RGB, depth only, or RGB-D images. We address those problems by proposing a novel object segmentation and detection system. Instance segmentation and object detection are significant problems in the fields of computer vision and robotics.
0 kommentar(er)
