Members: Sergio Casas Pastor, Sanket Biswas, Josep Brugués i Pujolràs
The tasks for the second week include:
-
Filter noise with linear or non-linear filters.
-
On denoised images, detect box with overlapping text, and apply OCR to get the text. Test query system using QSD1-W2 using only text.
-
Implement texture descriptors, and test them using QSD1-W2 using only texture descriptors.
-
Combine descriptors (text + color, text + texture, texture + color, text + color + texture) and test the retrieval on QSD1-W3.
-
Repeat the previous analysis with QSD2-W3 (remove noise, remove background, find 1 or 2 paintings and return the correspondences.
Slides for Week 3 are available here: Slides
main.py <querySetPath> <backgroundRemoval> <textRemoval> <textRemovalMethod> <k>
The querySetPath parameter indicates the location of the query set images.
The backgroundRemoval parameter indicates if background has to be removed from the query set images.
- True: background removal
- False: no background removal.
The textRemoval parameter indicates if the text has to be removed from the query set images.
- True: text removal
- False: no text removal.
The textRemovalMethod parameter indicates the method used for text detection:
- 1 : text detection based on color segmentation
- 2 : text detection based on morphology operations
- 3 : text detection based on EAST opencv neural network
The k parameter is used to indicate the number of top results that need to be saved.