Three-dimensional scene semantic completion method based on implicit representation
An implicit representation and 3D scene technology, applied in the field of semantic completion of 3D scenes, can solve the problems of inconvenient downstream applications and meet the needs of geometric variable granularity
Pending Publication Date: 2022-07-22
NANJING UNIV
View PDF0 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Therefore, in order to meet the needs of variable granularity, it is necessary to readjust the network structure and retrain each time, which is extremely inconvenient for some downstream applications.
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0080] like figure 1 As shown, the present invention discloses an implicit representation-based 3D scene semantic completion method, which is specifically implemented according to the following steps:
[0081] Step 1, build an implicit encoding and decoding network: build an implicit representation encoding network for generating implicit representations and an implicit representation decoding network for implicit representation decoding.
[0082] Step 2, latent representation semantic label generation: For each training sample, a certain number of points are randomly sampled in the three-dimensional space, and the category label of each point is calculated according to the voxel-level label, and the latent representation semantic label is generated.
[0083] Step 3, training of the hidden representation encoder-decoder network: According to the stochastic gradient descent algorithm, the hidden representation encoder-decoder network is trained by using the hidden representatio...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention discloses a three-dimensional scene semantic completion method based on implicit representation. The method comprises the following steps: step 1, implicit coding and decoding network construction: constructing an implicit representation coding network used for generating implicit representation and an implicit representation decoding network used for implicit representation decoding; 2, generating a hidden representation semantic tag; step 3, implicit representation coding and decoding network training: according to a stochastic gradient descent algorithm, utilizing implicit representation semantic tags to train an implicit representation coding and decoding network; and step 4, generating a voxel semantic completion result. According to the invention, through the direct connection between the decoupling granularity and the network structure, the effect of meeting the requirement of variable granularity on the premise of not adjusting the network and retraining is achieved. According to the method, the scene completion of any granularity can be realized, and the requirements of different application occasions on the geometric variable granularity of the scene model can be met.
Description
technical field [0001] The invention relates to a three-dimensional scene semantic completion method, in particular to a three-dimensional scene semantic completion method based on implicit representation. Background technique [0002] 3D scene semantic completion aims to recover complete scene geometry from incomplete scene geometry, and use semantic information to guide the completion process in the process. At present, this technology has been widely used in various 3D scene modeling occasions such as intelligent robots, virtual reality, and augmented reality. However, since different occasions have different requirements for the granularity of model geometry, how to dynamically adapt to the requirements of varying granularity of geometric models in different occasions has become a difficult problem in semantic completion of 3D scenes. [0003] According to the representation method of the scene 3D model, the scene semantic completion technology can be divided into the c...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Applications(China)
IPC IPC(8): G06T15/00G06N3/08G06N3/04
CPCG06T15/00G06N3/08G06N3/045
Inventor 孙正兴骆守桐孙蕴瀚王一
Owner NANJING UNIV