Summary In the recent years, there is a growing trend toward digitization of cultural heritage for better accessibility and preservation. For instance, the development of image processing techniques in traditional Chinese painting (TCP) has begun to attract researchers' attention in the computer vision field. TCP is one of the representative of Chinese traditional arts. Evidenced by the successes of development in image processing techniques in various applications, this article aim to apply the deep learning approach on TCP for several purposes, which include automatic establishment of unified image library, facilitating update‐to‐date data in the database, reduction of cost required for image classification and retrieval. First, a unified database is established, that consists of more than a thousand of images from six major TCP themes. Then, several deep learning algorithms that are based on mathematical models are applied to examine the classification performance. In addition, the salient regions that denote significant features are identified, by adopting the instance segmentation technique. As a result, the modified pretrained neural network is capable to achieve 99.66% recognition accuracy. Qualitative results are also presented to demonstrate the effectiveness of the proposed method. We also note that this is the first work that performs multiclass classification on six categories in this domain. Furthermore, a 10‐class classification result of 96% is obtained when performing on one of the painting types, namely, ghost‐and‐god.

In the recent year, state-of-the-art for facial micro-expression recognition have been significantly advanced by deep neural networks. The robustness of deep learning has yielded promising performance beyond that of traditional handcrafted approaches. Most works in literature emphasized on increasing the depth of networks and employing highly complex objective functions to learn more features. In this paper, we design a Shallow Triple Stream Three-dimensional CNN (STSTNet) that is computationally light whilst capable of extracting discriminative high level features and details of micro-expressions. The network learns from three optical flow features (i.e., optical strain, horizontal and vertical optical flow fields) computed based on the onset and apex frames of each video. Our experimental results demonstrate the effectiveness of the proposed STSTNet, which obtained an unweighted average recall rate of 0.7605 and unweighted F1-score of 0.7353 on the composite database consisting of 442 samples from the SMIC, CASME II and SAMM databases.