Abstract:
In last decade, With the rise of neural networks Depth completion has become extensive attention
recently due to the development of autonomous driving, which aims to recover dense depth
points from sparse depth measurements. In this thesis first introduces the rise and development
of deep learning and convolution neural network and summarizes the basic traditional methods,
and a brief background of depth estimation, pooling operation of convolution neural network
convolution feature extraction. Convolutional spatial propagation network (CSPN) is one of the
state-of-the-art methods and we are going to dig through all the methods to do a performance
comparison with the traditional methods. Then, the research status and development trend of
convolution neural network based on depth completion in sparse data are reviewed, which is
mainly introduced from the aspects of typical network structure construction, framework, training
method and performance. Finally, some problems in the current research are briefly summarized
and discussed. Later, we will conclude the literature review based on pros and cons od state-ofthe-art methods and how recent research has move forward for CNN with depth completion on
sparse data for current computer vision problems.
