PCA error using CV_PCA_DATA_AS_COL (Bug #701)
| Status: | Done | Start date: | |
|---|---|---|---|
| Priority: | High | Due date: | |
| Assignee: |  Vadim Pisarevsky |
% Done: | 0% |
| Category: | core | ||
| Target version: | - | ||
| Affected version: | Operating System: | ||
| Difficulty: | HW Platform: | ||
| Pull request: |
Description
When using PCA with CV_PCA_DATA_AS_COL as parameter there is an error in eigenvector matrix used for "project".
In my case I have 20 elements with 64 components compressed in 32.
I have an input matrix 64 rows x 20 columns
What I expect is and eigenvector matrix made of 64rows and 20 columns, instead it is 20 rows and 64 cols.
I think that the error is in PCA& PCA::operator() function where out_count is reduced to 20 instead of remaining 32!
Associated revisions
Merge pull request #701 from jet47:clahe
History
Updated by Vadim Pisarevsky over 14 years ago
CV_PCA_DATA_AS_COL/CV_PCA_DATA_AS_ROW is only used and known at the decomposition step. The only reason why CV_PCA_DATA_AS_COL is supported is because the data can be huge and it can be too expensive (time-wise and memory-wise) to transpose it. After PCA is performed, the project() and backProject() functions work on row-oriented data. That's why eigenvector matrix is always computed as if the data had CV_PCA_DATA_AS_ROW layout.
- Status changed from Open to Done
- (deleted custom field) set to worksforme