OPTIMOL: automatic Object Picture collecTion via Incremental MOdel Learning
description
Transcript of OPTIMOL: automatic Object Picture collecTion via Incremental MOdel Learning
OPTIMOL: automatic Object Picture collecTion
via Incremental MOdel Learning
a chicken and egg problem…
…among users, researchers, and data
ImagesImages
e.g. Caltech101,LabelMe, LHI
Framework
Category model
Classification
Dataset
Keyword: accordion Li, Wang & Fei-Fei, CVPR 2007
Framework
Category model
Classification
Dataset
Keyword: accordion Li, Wang & Fei-Fei, CVPR 2007
…
Compute SIFT descriptor
[Lowe’99]
Image representationImage representation
Kadir&Brady interest point detector
Codewords representatio
n
Nonparametric topic model-Hierarchical Dirichlet Process (HDP)
MN
Each patch
Each image
Teh, et al. 2004; Sudderth et al. CVPR 2006; Wang, Zhang & Fei-Fei, CVPR 2006
Nonparametric topic model-Hierarchical Dirichlet Process (HDP)
MN
Teh, et al. 2004; Sudderth et al. CVPR 2006; Wang, Zhang & Fei-Fei, CVPR 2006
Li, Wang & Fei-Fei, CVPR 2007
Classification
Likelihood ratio for decision:
Category likelihood for I:
Li, Wang & Fei-Fei, CVPR 2007
Annotation
Pitfall #1: model drift
Object Model
…
Object Model
…
Li, Wang & Fei-Fei, CVPR 2007
Object Model
Good Images
Bad Images
…
Pitfall #2: model diversity
Li, Wang & Fei-Fei, CVPR 2007
The “cache set”
Li, Wang & Fei-Fei, CVPR 2007
CategoryModel
classification
Enlarged dataset
Cache
Raw image dataset
Incrementallearning
Result
Li, Wang & Fei-Fei, CVPR 2007
Li, Wang & Fei-Fei, CVPR 2007
Li, Wang & Fei-Fei, CVPR 2007
OPTIMOL also learns good models
Team OPTIMOL (UIUC-Princeton):
11stst Place Place in the Software League