Fast Single-View 3D Object Reconstruction With Fine Details Through Dilated Downsample And Multi-Path Upsample Deep Neural Network

Three-dimensional (3D) object reconstruction is among the mostimportant research areas in the field of computer vision. Its pur-pose is to reconstruct the overall shape of an object from its two-dimensional (2D) image. With the development of deep learning,many methods based on convolutional neural networks (CNNs) havebeen applied in related research.To achieve 3D shape reconstruction with low computation time,we focus on the commonly used method: single-image reconstruc-tion. The main issue of using a single image as an input is that thereconstruction shape often lacks structural detail. To address this is-sue, we proposed two methods: the dilated downsample block andthe multi-path upsample block. The dilated downsample block ex-tracts more features and the multi-path upsample block uses the fea-tures in our architecture. Thereafter, we concatenate the encoder anddecoder with corresponding layers to keep the image features in re-construction process.Finally, we perform experiments on the dataset provided byChoyet al.Results show that our method achieves 67.7%intersection-over-union (IoU) accuracy, 3.6%higher than state-of-the-art method,VTN. Compared to the PSVH method, our result achieves 71.4%,an increase of 3.4%. Our average reconstruction time is 13 ms,approximately 25 times faster than PSVH.