1 / 24

Business Identification: Spatial Detection

Business Identification: Spatial Detection. Alexander Darino Week 7. Weaknesses to Current Approach. Business Name Matching. Business Spatial Detection. Latitude Longitude. Geocoding Reverse Geocoding. Nearby Businesses. Business Identification. Image. OCR. Detected Text.

rae
Download Presentation

Business Identification: Spatial Detection

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Business Identification:Spatial Detection Alexander Darino Week 7

  2. Weaknesses to Current Approach Business Name Matching Business Spatial Detection Latitude Longitude Geocoding Reverse Geocoding Nearby Businesses BusinessIdentification Image OCR Detected Text

  3. Alternative: Image Matching

  4. Alternative: Image Matching • Weaknesses: • Low Availability of Storefront Images (< 50% Avg) • George Aiken area businesses with photos: 18/35 • Brueggers area businesses with photos: 22/40 • Tambellini area businesses with photos: 8/22 • Available Images too small (100 x 100) • Not a viable solution

  5. Alternative: Template Matching • Tambellini • Tambellini • Tambellini • Tambellini • Tambellini • Tambellini • Tambellini • Tambellini

  6. Alternative: Template Matching • Progress • Able to generate templates • Able to extract SIFT features using Lowe’s implementation • Able to match SIFT features using Lowe’s implementation • Problems • Features are being matched to garbage

  7. Alternative: Template Matching

  8. Alternative: Template Matching

  9. Alternative: Template Matching

  10. Alternative: Template Matching

  11. Alternative: Template Matching • Currently on-hold • Need to discuss solution with Amir • Currently looking into another alternative…

  12. Alternative: Scene Text Recognition • State of the Art: • STR ≠ OCR • Far superior to our ‘naïve’ approaches to STR (ie. OCR, Image matching, SIFT) • OCR only works for highly controlled environments. CEDAR, ICDAR, etc not helpful • STR works for unconditioned environments • Scale invariant • Color/intensity invariant • Font invariant • Lexicon-Assisted

  13. Alternative: Scene Text Recognition • No STR implementations readily available • University of Massachusetts specializes in STR • Papers describe enhancements and unification of previous work, but not algorithms • Will email for blackbox implementation • Currently looking into ‘previous work’ • More models • Some algorithms

  14. Alternative: Scene Text Recognition • Options • Email authors for implementation • Try to implement STR as per described models • Blackboxes whenever possible (email!) • Code when blackboxes are not available • Try to implement crude STR via blackboxes Increase Contrast OCR Detected Text Text Detection Orthorectification

  15. STR Implementation • STR Implementation: “Automatic Detection and Recognition of Signs From Natural Scenes” Multiresolution-based potential characters detection Character/layout geometry and color properties analysis Refined Detection Local affine rectification

  16. Multiresolution-based potential characters detection • Laplacian-of-Guassian Edge Detection • Dice image/edges into Patches • Combine patches with similar properties into regions • Obtain bounding box of region as candidate text • Properties include: • Mean • Variance • Intensity(?)

  17. Multiresolution-based potential characters detection

  18. Multiresolution-based potential characters detection Patches qualify if:

  19. Multiresolution-based potential characters detection

  20. Multiresolution-based potential characters detection

  21. Multiresolution-based potential characters detection

  22. STR Implementation • Possible Solutions: • Don’t grow bounding box. Grow non-rectangular region, then obtain bounding box • Or replace with off-the-shelf Text Detector blackbox (?)

  23. Next Steps • Email for STR implementations • Backtrack: Implement ‘crude’ STR • Continue with current STR implementation Increase Contrast/Binarization OCR Detected Text Text Detection Orthorectification

  24. Thank You

More Related