Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
Upcoming SlideShare
Loading in …5
×

# slides

341 views

Published on

• Full Name
Comment goes here.

Are you sure you want to Yes No
Your message goes here
• Be the first to comment

• Be the first to like this

### slides

1. 1. Semantic Geometric Features: A Preliminary Investigation of Automobile Identification Carl E. Abrams Sung-Hyuk Cha, Michael Gargano, and Charles Tappert
2. 2. Agenda <ul><li>Overview of the Problem </li></ul><ul><li>The Experiments </li></ul><ul><li>Results </li></ul><ul><li>Going Forward </li></ul>
3. 3. Overview <ul><li>Object recognition remains a hard problem </li></ul><ul><li>The human mind uses shapes to recognize objects </li></ul><ul><li>Can semantic features defined by their shapes be more effective in the recognition and identification of objects? </li></ul>
4. 4. The Experiments <ul><li>10 test images of cars </li></ul><ul><li>Directly form the manufactures websites </li></ul><ul><li>Images were restricted to side views of the cars taken from 90 degrees </li></ul><ul><li>All 2005 models </li></ul><ul><li>Feature vectors calculated/measured from the images </li></ul>
5. 5. The Vehicles
6. 6. Experiments used Euclidean Distance as the Measure the xi and ti are measurements from two different vehicles
7. 7. Experiments used Euclidean Distance as the Measure (x1,y1) (x2,y2) c = (a 2 +b 2 ) 1/2 c = ((x1-x2) 2 +(y1-y2) 2 ) 1/2 c a b
8. 8. Manufacturers Specifications First Experiment
9. 9. Boundary Description using Rays Second Experiment
10. 10. Semantic Features Third Experiment
11. 11. Challenge: Determine the qualitative ability of the feature vectors to separate the vehicles <ul><li>Within each experiment compute the distance of each vehicle from all the others </li></ul><ul><li>Evenly divide the measures into 5 bins </li></ul><ul><li>Observe the distribution of the measures </li></ul>
12. 12. The Results
13. 13. Distance Matrix – Semantic Features Honda Civic Honda Accord Mazda 3 Mazda 6 Porsche Carerra Toyota Camry Toyota Celica Toyota Corolla Toyota Echo VW Passat Honda Civic Honda Accord Mazda 3 Mazda 6 Porsche Carerra Toyota Camry Toyota Celica Toyota Corolla Toyota Echo VW Passat 0 154.01 10.48 156.53 13.34 162.50 5.00 7.55 4.12 156.23 154.01 0 148.33 28.00 151.63 51.07 151.06 150.09 154.01 26.07 10.48 148.33 0 152.24 5.83 159.84 6.55 9.43 13,45 151.90 156.53 28.00 152.24 0 156.36 27.58 154.36 153.60 156.36 2.82 13.34 151.63 5.83 156.36 0 164.35 9.43 11.44 16.09 155.96 162.50 51.08 159.84 27.58 164/35 0 160.89 159.77 161.73 27.65 5.00 151.06 6.55 154.37 9.43 160.89 0 4.24 7.07 154.00 7.55 150.09 9.43 153.60 11.44 159.78 4.24 0 7.21 153.17 4.12 154.01 13.45 156.36 16.09 161.72 7.07 7.21 0 156.02 156.23 26.07 151.90 2.82 155.96 27.66 154.00 153.17 156.02 0
14. 14. Going Forward <ul><li>Extend techniques to encompass semantic shapes within an object (shape contexts) </li></ul><ul><li>Compare the extended semantic methods to existing methods in multiple domains </li></ul>
15. 15. Going Forward Shape Contexts
16. 16. References <ul><li>[1] R. D. Acqua and R. Job, &quot;Is global shape sufficient for automatic object identification?&quot; Congitive Science , vol. 8, pp. 801-821, 2001. </li></ul><ul><li>[2] A. K. Jain, A. Ross, and S. Pankanti, &quot;A Prototype Hand Geomtery-based Verification System,&quot; presented at Proceedings of 2nd International conference on Audio and Video-based Biometric Person Authentication, Wahington D.C., 1999. </li></ul><ul><li>[3] H. Schneiderman and T. Kanade, &quot;A Statistical Model for 3D Object Detection Applied to Faces and Cars,&quot; presented at IEEE Conference on Computer Vision and Pattern Recognition, IEEE, 2000 </li></ul><ul><li>[4] S. Belongie,J Malik, J Puzicha, “Matching Shapes” ,presented at the International Conference on Computer Vision (ICCV 01) Vol 1, Jan 2001 </li></ul>