|
|
How Large is Large Scale Image Search on Mobile Phones? |
|---|---|
| Objective | Evaluate the feasability of large scale image search approaches on mobile phones. |
| Description | In recent years, approaches for efficiently identifying a query image of an object in databases of millions of images emerged. Efficient image descriptions via clustering of local features descriptors into visual words and employing fast nearest neighbor search structures (e.g., via vocabulary trees) were developed to match corresponding images. Yet, as the size of the image database increases, so do the memory requirements for storing the search structure. Therefore, large scale image search is typically running on dedicated servers that can be queried by mobile phones through capturing and transmitting a query image. As this query is typically happening over mobile networks a delay of several seconds between the query and the answer can occur (as in Google Goggles). In certain application areas the search space of the images can be reduced from millions to tens of thousands or even more. Then conducting the search directly on the phone could become feasible. The goal of this project is to evaluate the limits of large scale image search on mobile phones given existing techniques fast nearest neighbor search for visual words.
Requirements
|
| Team | Computer Vision |
| Contact | Grubert Jens |
| Offered as |
Bachelor's thesis Semester project |
| Duration |
2011-Oct
-
2012-Mar
|
