On the evening of May 14, the 8th @ World "Touching Academic Frontiers" International Exchange Event was successfully held in the conference room at the corner of the 2nd floor of the SCS. As one of the important measures to help students broaden their international perspectives and learn the latest cutting-edge knowledge, this @ World "Touching the Academic Frontiers" activity continues to invite four outstanding students to share their experiences in participating in international conferences and short-term overseas study programs, and to discuss the forefront topics of their respective fields.
Zhang Yuqi shared her experience of attending ICCV23 and presented a paper on representation learning of video editing. Representation learning in video editing is a method of extracting and learning useful features from video data through deep learning algorithms, in order to better understand and process videos. The goal of representation learning is to transform high-dimensional and complex raw video data into low dimensional, easy to process feature representations. These feature representations can be used for various tasks, such as video classification, video retrieval, activity recognition, scene understanding, etc.
Yao Jixun delivered a report themed with "Noise Robust Speech Conversion with the Background Sound Preserved". The main idea of noise robust speech conversion with background sound preserved is to convert the source speech into the target speech in the presence of background noise, while maintaining consistency in the background noise. He proposed an end-to-end framework based on multitask learning, which achieves the function of preserving the background sound of the source speech after conversion by sequentially cascading the separation module, bottleneck feature extraction module, and speech conversion module of the source speech. The experiment shows that this scheme is significantly better than the baseline system in terms of speaker similarity and sound quality.
Zeng Xi introduced a method of designing biological sequences based on evolutionary reinforcement learning without any priori knowledge. In order to save experimental costs, machine learning methods are used to assist in the design of biological sequences. However, due to the limited availability of biological sequence data with known characteristics, this severely limits the training of machine learning models and greatly reduces their applicability and performance. To fill this data gap, Zeng Xi proposed ERLBioSeq, an evolutionary reinforcement learning algorithm for biological sequence design that can design biological sequences without relying on priori knowledge. The experiment shows that the proposed method has achieved significant improvements compared to existing methods.
Fan Mingkai shared his learning experience at the School of Information Technology at the United Arab Emirates University and briefly described small object detection in complex environments. Small object detection in complex environments is a challenging task in the field of computer vision. This task involves detecting and identifying smaller targets in images or videos with a large amount of background interference and complex scenes. Based on his own research experience, Fan focused on how to enhance the model's ability to represent small targets and enhance its detection ability for weak targets.
With the wonderful narration of the last sharer coming to an end, SCS's 8th @ World "Touching Academic Frontiers" International Exchange Activity has wrapped up successfully. This event not only provides students with a window to gain a deeper understanding of international conferences and short-term study visits, but also lays a solid foundation for future international exchanges and cooperation. It expands the international perspective of students and arouses their enthusiasm for scientific research exploration, injecting new impetus into their future academic journey.
In the future, the SCS will continue to focus on building a broader academic exchange platform for students and promoting the in-depth development of international education. Let's join together at the next @ World "Touching Academic Frontiers" international exchange event.
