My research areas include Computer Vision and Machine Learning. In particular, I am focusing on applying machine learning methods to solve computer vision problems such as Improving Image Retrieval Results and Image Understanding
I am also interested in image interpretation based on multiple cues and applying computer vision and natural language techniques to things on the internet.
Recently, I am interested in image represenation and the potential of high level image representation.
Projects:
 |
Object Bank Representation
Robust low-level image features have been proven to be effective representations for a variety of tasks such as object recognition and scene classification; but pixels,
or even local image patches, carry little semantic meanings. For high level visual tasks, such low-level image representations are potentially not enough. We propose a high-level image representation, called the Object Bank,
where an image is represented as a scale-invariant response map of a large number of pre-trained generic object detectors, blind to the testing dataset or visual task. As we try to tackle higher level visual recognition problems, we show that Object Bank representation
is powerful on scene classification tasks because it carries rich semantic level image information.
Full Text: (Parts&Attribute) PDF (NIPS) PDF |
|
Total Scene Understanding
Classification, annotation and segmentation are three challenging problems in high level recognition research. The good news is that they are mutually beneficial to each other. We propose a hierarchical generative model which takes into account this merit. Our framework simutanously classifies the overall scene, recognizes and segments each object component, as well as annotates the image with a list of tags.
Full Text: (CVPR09) PDF PPT Project Link |
 |
Understanding Event from Static Image
Based on the success of scene understanding and object recognition, we take a further step to interpret a static event image. Specifically, we are interested in the sports event understanding. We propose a first attempt to classify events in static images by integrating scene and object categorizations.
Full Text: (ICCV07) PDF |
|
OPTIMOL: automatic Object Picture collecTion via Incremental MOdel Learning
With its explosive growth, the Internet has become a key resource for users to obtain multi-media information such as images. Unfortunately, of the hundreds and thousands of images retrieved by the internet image search softwares, only a small fraction of them would be related images. How to use object recognition approach to help this out?
Related:
Full Text: (CVPR07)PDF (IJCV09)PDF Project Link
UIUC-Princeton Team in Semantic Robot Vision Challenge (AAAI 07)
|
|
