A Straightforward and Efficient Instance-Aware Curved Text Detector.

A challenging aspect of scene text detection is to handle curved texts. In order to avoid the tedious manual annotations for training curve text detector, and to overcome the limitation of regression-based text detectors to irregular text, we introduce straightforward and efficient instance-aware curved scene text detector, namely, look more than twice (LOMT), which makes the regression-based text detection results gradually change from loosely bounded box to compact polygon. LOMT mainly composes of curve text shape approximation module and component merging network. The shape approximation module uses a particle swarm optimization-based text shape approximation method (called PSO-TSA) to fine-tune the quadrilateral text detection results to fit the curved text. The component merging network merges incomplete text sub-parts of text instances into more complete polygon through instance awareness, called ICMN. Experiments on five text datasets demonstrate that our method not only achieves excellent performance but also has relatively high speed. Ablation experiments show that PSO-TSA can solve the text's shape optimization problem efficiently, and ICMN has a satisfactory merger effect.

Towards effective deep transfer via attentive feature alignment.

Training a deep convolutional network from scratch requires a large amount of labeled data, which however may not be available for many practical tasks. To alleviate the data burden, a practical approach is to adapt a pre-trained model learned on the large source domain to the target domain, but the performance can be limited when the source and target domain data distributions have large differences. Some recent works attempt to alleviate this issue by imposing feature alignment over the intermediate feature maps between the source and target networks. However, for a source model, many of the channels/spatial-features for each layer can be irrelevant to the target task. Thus, directly applying feature alignment may not achieve promising performance. In this paper, we propose an Attentive Feature Alignment (AFA) method for effective domain knowledge transfer by identifying and attending on the relevant channels and spatial features between two domains. To this end, we devise two learnable attentive modules at both the channel and spatial levels. We then sequentially perform attentive spatial- and channel-level feature alignments between the source and target networks, in which the target model and attentive module are learned simultaneously. Moreover, we theoretically analyze the generalization performance of our method, which confirms its superiority to existing methods. Extensive experiments on both image classification and face recognition demonstrate the effectiveness of our method. The source code and the pre-trained models are available at https://github.com/xiezheng-cs/AFAhttps://github.com/xiezheng-cs/AFA.