CV | Dong Liang’s Blog

邮箱
cs_liangdong@qq.com

电话
+86-17774010899

技能
Deep Learning, Machine Learning

框架
Tensorflow, Keras

Liang, Dong, et al. Unsupervised Teacher-Student Model for Large-Scale Video Retrieval." Proceedings of the IEEE International Conference on Computer Vision Workshops. 2019.Paper,Code
Liang D, Lin L, Chen X, et al. Multi-Stream Scale-Insensitive Convolutional and Recurrent Neural Networks for Liver Tumor Detection in Dynamic Ct Images[C]//2019 IEEE International Conference on Image Processing (ICIP). IEEE, 2019: 794-798.Paper, Code
Chen X, Lin L, Liang D, et al. A Dual-Attention Dilated Residual Network for Liver Lesion Classification and Localization on CT Images[C]//2019 IEEE International Conference on Image Processing (ICIP). IEEE, 2019: 235-239.Paper, Code
Liang D, Lin L, Hu H, et al. Combining convolutional and recurrent neural networks for classification of focal liver lesions in multi-phase CT images[C]//International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, Cham, 2018: 666-675.Paper, Code
Liang D, Lin L, Hu H, et al. Residual convolutional neural networks with global and local pathways for classification of focal liver lesions[C]//Pacific Rim International Conference on Artificial Intelligence. Springer, Cham, 2018: 617-628.Paper, Code

我是负责相似视频检索研发的实习生。

开发了无监督的Teacher-Student模型，来优化大规模相似视频检索。并且达到的state-of-the-art在FIVR数据集上，在三个任务上的mAP分别达到 50.94%, 49.83, 43.18. QPS(单卡P4) 提升两倍 (150->300).
基于hash的二值化特征，开发了帧级别的检索流程。它适合长视频检索的场景，可以保持视频的关键信息。
对比了不同检索引擎的性能。例如 HNSW, IVF-HNSW, Link && Code
相关工作已经发表至ICCV 2019 Workshop Paper,Code

Last Updated: 10/11/2019