We are happy to announce that Chia-Ming Li and Ming-Shyen Wu from ACVLab, National Cheng Kung University, stationed in Paris, have achieved remarkable success at the VIPriors challenge during IEEE ICCV 2023! Competing against 178 teams from around the world, they emerged victoriously, winning the unique Jury Prize.
This prestigious competition seeks high performance and values creativity and academic significance. The core aim of the competition is to achieve stable and exemplary performances in classic computer vision tasks, such as Instance Segmentation and Object Detection, with limited training data.
The VIPriors challenge underscores the importance of achieving superior results with small models and minimal GPU usage, aligning with global sustainability goals. The competition does not allow the use of pre-trained models, and all training must be conducted on extremely limited provided data, fostering innovation in Few-shot Learning methods.
Our team, especially Mr. Li, has dedicated immense time and effort and has designed highly innovative solutions. We introduced a method called the Memory-efficient Pipeline. This approach requires only a single NVIDIA RTX 2080 graphics card for training and avoids extensive Test-time Augmentation during inference. With this innovative approach, we not only reached the ultimate ranking of Top 3% but also won the Jury Prize for the originality of the method. This meaningful and honorable award strongly manifests our research capability and innovative spirit.
—–
*IEEE ICCV（International Conference on Computer Vision）是AI和計算機視覺（CV）領域的頂尖研討會之一，每次會議都會吸引全球的學者和專家來交流最新的研究成果和技術。它被廣泛認為是推動計算機視覺技術和理論發展的重要平台。
