Abstract: The present disclosure provides a point autoencoder including an encoder receiving a 3D point cloud and extracting a global feature and a decoder reconstructing the 3D point cloud from the global feature; a dual autoencoder including the point autoencoder; and a method for dimensional transformation of a point cloud using the dual autoencoder. According to the present disclosure, it is possible to extract features from an input point cloud using a Bias-Induced Decoding (BID) layer, perform reconstruction of a 3D point cloud from the corresponding feature information, and reconstruct a 3D point cloud by using only high-dimensional global features as an input. Also, the present disclosure reduces the amount of storage space used and reduces computational complexity than the conventional structure employing a fully connected layer.

Abstract: A scene flow estimation apparatus includes: a hierarchical feature detector detecting a hierarchical feature of point cloud data related to an image frame; a corrector correcting the detected hierarchical feature of the point cloud data; a bi-directional feature detector detecting a bi-directional scene flow feature of the corrected point cloud data, re-inputting a detection value, and repeatedly detecting the bi-directional scene flow feature; and an inferrer inferring a next scene flow of the image frame by using the bi-directional scene flow feature.

Abstract: An apparatus for distilling knowledge for a scene flow prediction model includes: a student model former forming a student model to have single bidirectional flow embedding and a flow predictor structure of a teacher model; a weight generator generating a weight based on a plurality of hierarchical prediction results of the teacher model and predetermined ground truth data; a function generator generating a loss function by using the weight and the plurality of prediction results; and a knowledge distiller distilling the knowledge of the teacher model to the student model by using the loss function.

Abstract: The present disclosure provides a point autoencoder including an encoder receiving a 3D point cloud and extracting a global feature and a decoder reconstructing the 3D point cloud from the global feature; a dual autoencoder including the point autoencoder; and a method for dimensional transformation of a point cloud using the dual autoencoder. According to the present disclosure, it is possible to extract features from an input point cloud using a Bias-Induced Decoding (BID) layer, perform reconstruction of a 3D point cloud from the corresponding feature information, and reconstruct a 3D point cloud by using only high-dimensional global features as an input. Also, the present disclosure reduces the amount of storage space used and reduces computational complexity than the conventional structure employing a fully connected layer.