彦究院

论文：

Guohua Cheng*, Jingliang Cheng, Mengyan Luo, Linyang He, Yan Tian, and Ruili Wang, “Effective and Efficient Multitask Learning for Brain Tumor Segmentation,” Journal of Real-time Image Processing, 2020. (SCI Q3, IF=1.96)

一种二分层图像分割方案，达到效果和效率的最佳平衡。

Guohua Cheng*, Hongli Ji, Yan Tian, “Walking on Two Legs: Learning Image Segmentation with Noisy Labels”, Uncertainty in Artificial Intelligence (UAI), 2020. (Oral, CCF-B, 142/515=27.5%)

针对标定数据中的噪声问题，基于GAN网络联合学习图像分割任务与噪声去除任务，副产品为样本学习的权重。

---

–

论文：

针对数据标定复杂问题，提出一种利用跨图像间辨识信息的弱监督学习方法。

Ruizhang, Yan Tian, Dongsheng Liu*. Weakly Supervised Learning to Detect Occluded Items in X-ray Baggage Inspection. Neurocomputing. (under review)

Yan Tian, Zhaocheng Xu, Guotang Jian, Xun Wang, Dongsheng Liu*. Weakly Supervised Learning for Detecting Occluded Items in X-ray Baggage Inspection. IEEE Trans on Multimedia. (under review)

专利：

徐照程, 田彦. 一种基于多尺度跨图像弱监督学习的违禁物检测方法. 专利申请号2021102756403, 申请日2021.03.15.