17-04-2012, 11:26 AM
Large Scale Image Processing with Hadoop
Large_Scale_Image_Processing.ppt (Size: 2.88 MB / Downloads: 129)
Big Data in Vision
Traditional Vision: Focus on the model
Pose Est.: 2D Image -> Virtual 3D model + Camera
Under-constrained, slow, sensitive to noise
Object Recognition: SVM + features
Breaks with many classes (e.g., every flickr tag)
New Trend: Focus on the data
DB of images (w/ metadata) -> query image
Problem becomes similar image search
Transfer metadata from DB images to query image
KNN methods simple and scalable
Clustering, hashing, metric learning
Hadoop+CV: Expectation Maximization
Map: Fit data to model given parameters (E-Step)
Red: Compute new model parameters given data (M-Step)
Iterate until stopping conditions are met.
Examples
Clustering (e.g., K-Means)
Mixture Models (e.g., MoG)
Image Retrieval with Hadoop
Analogies between image and text retrieval
Bag of Words -> Bag of Features
Document -> Image
Visual Word: Cluster of similar visual features
Compute Local Image Features (e.g., SIFT)
Cluster Features (i.e., create visual words)
Find cluster medians
Make Hamming Embeddings (compact feature) [1]
Efficient binary code (256 -> 8 Bytes per feature)
Hamming Distance
Benefit: Small size means more in memory
Inverted Index
Conclusion
Vision has 'Big Data' applications
Many image search applications
Common design patterns for M/R+Vision
Hadoop useful image search