Multi-view deep neural network is perhaps the most successful approach in 3D
shape classification. However, the fusion of multi-view features based on max
or average pooling lacks a view selection mechanism, limiting its application
in, e.g., multi-view active object recognition by a robot. This paper presents
VERAM, a recurrent attention model capable of actively selecting a sequence of
views for highly accurate 3D shape classification. VERAM addresses an important
issue commonly found in existing attention-based models, i.e., the unbalanced
training of the subnetworks corresponding to next view estimation and shape
classification. The classification subnetwork is easily overfitted while the
view estimation one is usually poorly trained, leading to a suboptimal
classification performance. This is surmounted by three essential
view-enhancement strategies: 1) enhancing the information flow of gradient
backpropagation for the view estimation subnetwork, 2) devising a highly
informative reward function for the reinforcement training of view estimation
and 3) formulating a novel loss function that explicitly circumvents view
duplication. Taking grayscale image as input and AlexNet as CNN architecture,
VERAM with 9 views achieves instance-level and class-level accuracy of 95:5%
and 95:3% on ModelNet10, 93:7% and 92:1% on ModelNet40, both are the
state-of-the-art performance under the same number of views.

本文提出了一种名为 VERAM 的递归注意模型，能够主动选择一系列视角进行高精度的三维形状分类，通过三个关键增强策略解决了现有注意力模型中不平衡训练的问题，实现了状态下的最先进水平。