Veggie Vision by IBM

1. Ideas about a practical system to make more efficient the selling and inventory of produce in a grocery store. Veggie Vision by IBM CSE 803 Fall 2013

2. Problem is recognizing produce • properly charge customer • do inventory • save customer and checker time CSE 803 Fall 2013

3. CSE 803 Fall 2013 15+ years of R&D now This information was shared by IBM researchers. Since that time, the system has been tested in small markets and has been modified according to that experience.

4. CSE 803 Fall 2013 Up to 400 produce types

5. CSE 803 Fall 2013

6. CSE 803 Fall 2013 Practical problems of application environment

7. Engineering the solution CSE 803 Fall 2013

8. CSE 803 Fall 2013 System to operate inside the usual checkout station • together with bar code scanner • together with scale • together with accounting • together with inventory • together with employee • within typical store environment * figure shows system asking for help from the cashier in making final decision on touch screen

9. CSE 803 Fall 2013 Modifying the scale

10. CSE 803 Fall 2013 Need careful lighting engineering

11. CSE 803 Fall 2013 Need to segment product from background, even through plastic

12. CSE 803 Fall 2013 Previously published thresholding decision

13. Quality segmented image obtained CSE 803 Fall 2013

14. CSE 803 Fall 2013 Design of pattern recognition paradigm (from 1997) FEATURES are: color, texture, shape, and size all represented uniformly by HISTOGRAMS Histograms capture statistical properties of regions – any number of regions.

15. CSE 803 Fall 2013 Matching procedure Sample product represented by concatenated histograms: about 400 D 350 produce items x 10 samples = 3500 feature vectors of 400D each Have about 2 seconds to compare an unknown sample to 3500 stored samples (3500 dot products) Analyze the k nearest: if closest 2 are from one class, recognize that class (sure)

16. CSE 803 Fall 2013 HSI for pixel color: 6 bits for hue, 5 for saturation and intensity For each pixel quantify H HIST[H]++ same for S&I

17. CSE 803 Fall 2013 Histograms of 2 limes versus 3 lemons Distribution or population concept adds robustness: • to size of objects • to number of objects • to small variations of color (texture, shape, size)

18. CSE 803 Fall 2013 Texture: histogram results of LOG filter[s] on produce pixels Leafy produce B Leafy produce A

19. CSE 803 Fall 2013 Shape: histogram of curvature of boundary of produce

20. CSE 803 Fall 2013 Banana versus lemon or cucumber versus lime Small range of curvatures indicates roundish object Large range of curvatures indicates complex object

21. CSE 803 Fall 2013 Size is also represented by a histogram

22. CSE 803 Fall 2013 Learning and adaptation System “easy” to train: show it produce samples and tell it the labels. During service: age out oldest sample; replace last used sample with newly identified one. When multiple labeled samples match the unknown, system asks cashier to select from the possible choices.

23. CSE 803 Fall 2013 Where is Veggie Vision today? http://www.internetnews.com/xSP/article.php/3642386 System uses almost all color features Installed in few places: many stores have self-checkout, putting work on customer. IBM has a “shopping research” unit http://www.usatoday.com/tech/news/techinnovations/2003-09-26-future-grocery-shop_x.htm Customers will tolerate a higher human error than a machine error