The Picasa Name Game
Source: Tom's Guide US | Keywords: facial, recognition, tagging | Themes: Software
- 1. The Tagging Challenge
- 2. Apple iPhoto ‘09
- 3. Apple iPhoto ‘09: First iPhoto Steps
- 4. Apple iPhoto ‘09: Wrinkles in Faces
- 5. Apple iPhoto ‘09: Funny Faces
- 6. iPhoto Analysis
- 7. CyberLink MediaShow 5
- 8. MediaShow 5: Warning--Wide Load
- 9. MediaShow 5: UI and Criticisms
- 10. MediaShow 5: Criticisms, Cont'd
- 11. MediaShow 5: Analysis
- 12. Face.com Photo Tagger
- 13. Photo Tagger: Opening Accuracy
- 14. Photo Tagger: The Tagging Process
- 15. Photo Tagger: Working With Photo Finder
- 16. Photo Tagger Analysis
- 17. Google Picasa 3.5
- 18. The Picasa Name Game
- 19. Picasa’s No Dog
- 20. Picasa Analysis
- 21. Microsoft Windows Live Photo Gallery
- 22. WLPG: Name On!
- 23. WLPG: Naming, Bonus Round
- 24. WLPG: Analysis
- 25. More on this topic
18. The Picasa Name Game
Keep in mind that Picasa is the one of the few face taggers to offer any level of analysis threshold control. You can look under Tools > Options > Name Tags tab to find sliders for Suggestion Threshold and Cluster Threshold. I left both of these at the default value of 80 for testing, but consider adjusting these if you’re not happy with your recognition results.
In my case, I highlighted the area of the left bar showing Unnamed (485). This brings up a gallery view of all recognized faces. For each of my four subjects, I picked one decent, straight-on face to see how Picasa would fare on its first recognition. Adding a name to a face brings up a People dialog box in which you can add contact information about the person and coordinate with your online Google contacts. Just to be quick rather than thorough, I only used first names and left all other fields blank. When ready to move on, click New Person. Alternatively, if you already have a contact made for this person, you can select it from the auto-fill name list that appears and hit the Choose button.
By this point, I knew exactly how many total face instances there were of these people in my test collection. Picasa’s opening results broke down like so: Me, 4 of 45, 100% accurate; Knico, 13 of 43, 100% accurate; Devon, 26 of 117, 100% accurate; Garrett, 26 of 93, 92% accurate. So out of 69 returns, 67 were positive matches. That’s freakin’ awesome! Even better, I noticed that when I clicked the red X icon under Garrett’s two false positives, they correctly reappeared under a fresh set of suggested faces for Devon. Picasa’s learning was outstanding. The second set of results for Devon had 14 of 17 faces correct. Upon clicking the Confirm all group for these 14, a third group appeared in which 5 out of 5 were correct. So in about 10 to 15 seconds, I already had 53 of 117 tags for Devon finished.
Moreover, I quickly realized that Picasa had implemented some of the asset mobility I’d wanted to see in MediaShow. For example, at one point, I discovered that I’d accidentally approved a person under the wrong name. No problem! You simply grab the mis-tagged face and drop it on the correct name under the People bar.
- Previous page Google Picasa 3.5
- Next page Picasa’s No Dog
Please update more on:
(.
1. let just say I have spend weekly time on a low end pc to get 33 GB photo with so many file. How do I save this tag when I am reinstall windows?
2. About the portabilty in no 1. Picasa has picasa.ini in every folder, but when it corrupt, the picasa.ini is not helpful recovering the weekly time spent.
3. Speed? Why there is no benchmark graph like usual?
4. Try gradiation photos or something similiar. It will see about the acuration.
5. I want to get the best speed, what is the most needed hardware. If Processor will I7 better than C2D? If GPU, will Geforce GTX 295 better than 9800?
I have private paint experience using picasa. I have taging many face in a week of Sempron 2800+ OC to 2 Ghz. When the face recognation is done, for what ever reason, my cpu is dead (dead power electricity). When the electricity power is up, my pc is on windows. The picasa is corrupt. My one week OC is for nothing. DAMN
I recently tagged all my photos with Picasa. I think i personally tagged more than the software did--the software is VERY cautious. It also repeatedly asked to tag posters, even paintings that were on my walls. Finally, it doesn't do well with babies--which is no surprise because they all look the same to me too :-)
Its cool, but im not sure the outcome was worth sitting there tagging hundreds of pictures of ex-girlfriends.
I am giving comment for testing.
Deadlockedworld--you don't have to tag everyone in every album in Picasa--just don't tag albums that have your exes in them. I definitely don't tag everyone in my photos--only those people that are important to me.
Adobe's Photoshop Elements 7 has the ability to detect faces but not automatically match them. It did make tagging much quicker. I could select 40 pictures from a party and tag the lot all at once.
However, I found that there were several pictures that it didn't catch. So, I ended up having to go through the whole bunch manually anyway to catch the stragglers. I found I spent as much time, if not more, making sure I got everything. So, I'm not sure that the "helpful tool" actually did much.
I am using iPhoto '09 and I found that the predictive tagging is getting less and less accurate as the database of tagged faces increases. Impression confirmed by one of my friends using the same app. For example, my wife is probably the most frequent face in my collection and the software has a hard time identifying her. On the other hand, I tagged the face of a friend I see rarely and I was welcomed with 4-5 good matches.