My Projects:
“Multi-modal Emotion Recognition with Graph Neural Networks” (intend to submit to AAAI 2021) Applied a novel hierarchical multi-modal feature fusion to the graph level. Used attention-based encoder with Graph Convolutional Networks. Trained using Valence-Arousal learning pipeline. Achieved new state-of-the-art results in the following data: IEMOCAP, MELD, AVEC. Library used: NetworkX, Pytorch Geometric. Supervised and granted funding from Prof. Hiroshi Ishiguro.
“Joint Fovea and Optic Disc Segmentation for Degenerated Retina” (submitted to MICCAI 2021) Proposed a new U-Net based attention model with multiple local bottleneck structures. Conducted extensive experiments and ablation studies to prove it in segmenting the landmarks in disease fundus images(Age-Related Macular Degeneration, Glaucoma), which are usually hard to segment with traditional methods. Achieved new state-of-the-art results in the following data: REFUGEE, Baidu ADMD, Messidor.
“Joint Modeling of EEG, fMRI and Structural MRI” (intend to submit to AAAI 2021) Combined high temporal resolution EEG and high spacial resolution fMRI for better brain activity modeling. Used spatial-temporal graph neural networks with adaptive adjacency matrix to solve the time series prediction. Interpreted the signal source localization among various brain tasks.
“Kaggle Competitions” I achieved top 530/150000 and Competition Expert tier in Kaggle by the end of 2020. This is the project repo for my competitions (4 silver medals and 2 solo bronze medal).
Updating… Thanks for reading!