• 

• Catherine Qi Zhao

  Associate Professor
  Department of Computer Science and Engineering
  University of Minnesota
  
  Office: Keller 5-213
  Phone: (612) 301-2115
  Email: qzhao at cs.umn.edu
  

• My research is in the areas of computer vision, machine learning, and healthcare. In particular, I am interested in providing theoretical foundations and computational innovations in computer vision, inventing new machine learning methods inspired by AI problems, and building intelligent systems that leverage both artificial and human intelligence.

  My recent focus includes understanding neural networks, and developing explainable and trustworthy AI models. We also apply these techniques to emerging healthcare applications, for example, to understand and identify neurodevelopmental disorders, and to decode human motor intention in upper limb amputees.

News

• our work on explicit knowledge incorporation in visual reasoning is out. [project page]
• our new work on predicting visual scanpath is out. [project page]
• I will be a program co-chair for WACV 2022.
• our work on attention and reasoning to quantify and improve the decision-making process is out. [project page]
• our work on leveraging attention as an interface to understand task performance is out. [project page]
• we receive an NSF RI grant to study attention and reasoning. [link]
• our work on autism screening with multi-modality information is out. [project page]
• I will be an area chair for CVPR 2020 and doctoral consortium chair for WACV 2020 and CVPR 2023.
• we receive an NSF S&AS grant to develop intelligent UAVs with active vision. [link]
• our work on shallowing deep neural networks is out. [project page]
• I will be an area chair for WACV 2019, CVPR 2019, and IJCAI 2019.
• our work on emotion and attention is out. [project page]
• we receive an NSF SHF grant to develop efficient time-based deep neural networks. [link]
• we will organize the 3rd LSUN Saliency Challenge, in conjunction with CVPR.
• the new book I edited is out -- it provides an overview of vision from various perspectives, ranging from neuroscience to cognition, and from computational principles to engineering.
  
  
  
• our work on autism photo is out in Current Biology.
• I have joined the University of Minnesota Twin Cities as an assistant professor.
• commentary about our autism work appears in Neuron.
• press release at HuffPost, Business Insider, Daily Mail, MedicalXpress, and Futurity.
• our work is on the cover of Neuron. [pdf]
  

More

• Database

  • AiR-D database. A database with attention and reasoning labels as well as ground truth answer correctness. Chen et al. ECCV [pdf] [bib]
  • IQVA database. A large-scale attention database on 360° videos with ground truth answer correctness. Jiang et al. CVPR [pdf] [bib]
  • SALICON database. Saliency in Context - a large-scale attention database on MS COCO images. Jiang et al. CVPR [pdf] [bib]
  • Video Story database. A dataset for generating text story/summarization for videos containing social events. Li et al. TMM [pdf] [bib]
  • EMOd database. A database with rich sentiment and semantic attributes. Fan et al. CVPR [pdf] [bib]
  • HII database. A dataset of images with four types of interactions: hand shake, high five, hug, kiss. Li et al. ACMMM [pdf] [bib]
  • PISC database. An image dataset consisting of 22,670 images of 9 types of social relationships. Li et al. ICCV [pdf] [bib]
  • OSIE database. Object and Semantic Images and Eye-tracking database - a database for object and semantic saliency (700 images, 5551 objects with fine contour and semantic attribute labeling). Xu et al. JoV [pdf] [bib]
  • EyeCrowd database. Eye Fixations in Crowd database - a database for saliency in crowd. Jiang et al. ECCV [pdf] [bib]

  More

• Code

  • Explicit Knowledge Incorporation for Visual Reasoning. Code for performing explainable visual reasoning with neural modules and explict knowledge incorporation.
  • Predicting Human Scanpaths in Visual Question Answering. Code for predicting human visual scanpaths in visual question answering, free-viewing, and visual search tasks.
  • AiR: Attention with Reasoning Capability. Code for implementing attention and reasoning supervision on the AiR-D dataset.
  • Immersive Question-directed Visual Attention. Code for implementing correctness-aware attention prediction on the IQVA dataset.
  • Direction Concentration Learning. Code for enhancing congruency (i.e., agreement between the learned knowledge and the new information in a learning process) in classification and continual learning.
  • Learning to Learn from Noisy Labeled Data. Code for performing noise-tolerant learning.
  • Attention-based Autism Spectrum Disorder Screening with Privileged Modality. Code for performing ASD screening on the publicly available Saliency4ASD dataset.
  • Context-Adaptive Saliency Network. Code for network implementation of context-adaptive saliency prediction, and evaluation metrics.

  More