A5101911851- Video Analysis and Summarization
- Generative Adversarial Networks and Image Synthesis
- Multimodal Machine Learning Applications
- Advanced Image and Video Retrieval Techniques
- Computer Graphics and Visualization Techniques
- Image Retrieval and Classification Techniques
- Music and Audio Processing
- Traffic control and management
- Data Mining Algorithms and Applications
- Digital Media Forensic Detection
- Handwritten Text Recognition Techniques
- Autonomous Vehicle Technology and Safety
- Data Quality and Management
- Vehicle License Plate Recognition
- Transportation and Mobility Innovations
- Big Data and Business Intelligence
- Advanced Image Processing Techniques
Wuhan University of Technology
2022-2024
Dalian University of Technology
2019
Automatic few-shot font generation (AFFG), aiming at generating new fonts with only a few glyph references, reduces the labor cost of manually designing fonts. However, traditional AFFG paradigm style-content disentanglement cannot capture diverse local details different So, many component-based approaches are proposed to tackle this problem. The issue is that they usually require special pre-defined components, e.g., strokes and radicals, which infeasible for languages. In paper, we present...
Different from focused texts present in natural images, which are captured with user's intention and intervention, incidental usually exhibit much more diversity, variability complexity, thus posing significant difficulties challenges for scene text detection recognition algorithms. The ICDAR 2015 Robust Reading Competition Challenge 4 was launched to assess the performance of existing methods on as well stimulate novel ideas solutions. This report is dedicated briefly introduce our...
Vehicle path planning problems have been studied for decades. The existing methods are suitable simple objectives. However, complex tasks such as paths vehicles considering the effects of pedestrians, traffic lights, etc., it is difficult to design a reasonable cost function deterministic algorithm or heuristic algorithm. In this paper, we proposes model based on light status and condition awareness. When vehicle arrives at new road section, senses status, distribution positions in network...
Text style transfer aims to the reference of one text image another image. Previous works have only been able a binary In this paper, we propose framework disentangle images into three factors: content, font, and features, then remix factors different new style. Both input no restrictions. Adversarial training through multi-factor cross recognition is adopted in network for better feature disentanglement representation. To decompose disentangled representation with swappable factors, trained...
Text-to-image retrieval, one of the most important cross-modality tasks, aims to search relevant images through a given text query. Most recent approaches are based on large-scale models. The huge time costs make it impossible for real-time searching. They also ignore fine-grained information, i.e., scene text. To tackle these issues, we propose novel matching method that considers in both modalities and adopts fast way by aligning from objects relations finally global. This logically...
Automatic few-shot font generation (AFFG), aiming at generating new fonts with only a few glyph references, reduces the labor cost of manually designing fonts. However, traditional AFFG paradigm style-content disentanglement cannot capture diverse local details different So, many component-based approaches are proposed to tackle this problem. The issue is that they usually require special pre-defined components, e.g., strokes and radicals, which infeasible for languages. In paper, we present...