高级检索

判别性特征引导的零样本三维模型分类算法

Discriminative Feature-Guided Zero-Shot Learning of 3D Model Classification Algorithm

  • 摘要: 基于零样本学习的三维模型分类是三维视觉领域的一个新兴话题, 旨在对未经训练的三维模型进行正确分类. 针对零样本三维模型分类中存在重视全局而忽视局部, 强制约束而无视语义-视觉跨域差异性, 导致整体性能低下的问题, 提出一种判别性特征引导的零样本三维模型分类算法. 首先以三维模型的多视图表征为输入, 自适应地捕获三维模型的局部判别性特征, 获得具有良好语义对应性的视觉特征表示; 然后以词向量的形式引入类的语义表示, 结合条件生成对抗网络生成类的伪视觉特征; 最后提出语义判别损失和内容感知损失联合监督, 从语义到内容共同约束真实视觉特征和伪视觉特征的对齐, 鼓励模型学习具有高局部判别性的特征, 实现语义-视觉的跨域细粒度对齐. 在ZS3D数据集上达到了60.9%的Top-1准确率, 超越当前最好方法2.3%的准确率, 同时在Ali的三个子数据集上也分别取得31.9%, 9.9%和16.6%的准确率, 均达到较好的实验效果, 验证了所提算法的有效性和普适性.

     

    Abstract:  Zero-shot learning of 3D model classification is a burgeoning topic in the field of 3D vision, aiming to classify untrained 3D models correctly. Aiming at the problem that zero-shot learning of 3D model classification focus on global information rather than local information, impose mandatory constraints, and ignore the cross-domains semantic-visual differences, resulting in low performance, this paper proposes a discriminative feature-guided zero-short 3D model classification network. Firstly, the local discriminative features, i.e., real-visual features of the multi-view 3D models, are adaptively captured by the proposed visual feature extraction module. Secondly, the semantic representations of the class labels are introduced in the form of word vectors, and their pseudo-visual features are generated by conditional generation adversarial network. Finally, the fine-grained across domains alignment of semantic-visual features is achieved by a novel semantic-content joint loss, which consists of semantic discriminative loss and content-aware loss between real-visual and pseudo-visual features from semantics to contents. The proposed algorithm achieves a Top-1 accuracy rate of 60.9% on the ZS3D dataset, which exceeds with the current best method with an accuracy rate of 2.3%. and achieves an accuracy rate of 31.9%, 9.9% and 16.6% on the three sub-datasets of Ali, respectively, which performs an excellent experimental result and verifies the effectiveness and universality of the proposed algorithm.

     

/

返回文章
返回