Sunday, June 17, 2007
A view from the palace
Monday, March 26, 2007
Improving CustomEyes
GlassesDirect offer a feature they call CustomEyes which allows you to "try on" the glasses they offer online. It works by you uploading a photo of your face, and then you can overlay an image of the glasses on top of that.
Unfortunately they don't really use the power of the computer in doing this. CustomEyes is a silly little Flash application, whereas this could all have been done in far more accessible HTML. You currently get to see a small image of your face, and then have to select the frame you want to see by name from a drop down. The Flash controls do allow you to rotate the glasses - but who wants to see glasses that are rotated off of the horizontal anyway? In addition, they also allow the glasses to be resized, which is necessary since you they don't know in advance what sized face image you have uploaded - but is doubly necessary because they don't even make all the glasses images the same size - some are twice the size of others!
I'd suggest that the application could be much improved by doing the following:
Unfortunately they don't really use the power of the computer in doing this. CustomEyes is a silly little Flash application, whereas this could all have been done in far more accessible HTML. You currently get to see a small image of your face, and then have to select the frame you want to see by name from a drop down. The Flash controls do allow you to rotate the glasses - but who wants to see glasses that are rotated off of the horizontal anyway? In addition, they also allow the glasses to be resized, which is necessary since you they don't know in advance what sized face image you have uploaded - but is doubly necessary because they don't even make all the glasses images the same size - some are twice the size of others!
I'd suggest that the application could be much improved by doing the following:
- do away with the Flash, and make this a straightforward dynamic HTML page
- select from an illustrated list of frames, rather than simply by text name
- for each frame that you select, add a new copy of the uploaded image to the page, so that you can actually fill the page with images that you can compare side by side, or which you can print out and compare that way
- either automatically recognise where the eyes are in the image when it is uploaded, or ask to be told once, and thereafter make sure that all the images of the frames are scaled to be the correct size to fit.
Wednesday, March 07, 2007
What are Zooomr up to this week?
Just setting things up so that I'm all ready to try out the new Zooomr version 3 when it's ready (which Thomas Hawk is saying is very soon now).
moon Hosted on Zooomr
moon Hosted on ZooomrLabels: zooomr
Wednesday, February 14, 2007
YouTube blog
YouTube has millions of users.
It also has a blog at http://youtube.com/rss/global/our_blog.rss
According to Bloglines, there are just 9 subscribers to that blog via Bloglines!
I wonder, is it that YouTube users are not very blog savvy, that the YouTube blog is not very interesting, or are YouTube blog readers not doing it via Bloglines?
(By comparison, the official Google Blog reports 43,697 subscribers via Bloglines).
It also has a blog at http://youtube.com/rss/global/our_blog.rss
According to Bloglines, there are just 9 subscribers to that blog via Bloglines!
I wonder, is it that YouTube users are not very blog savvy, that the YouTube blog is not very interesting, or are YouTube blog readers not doing it via Bloglines?
(By comparison, the official Google Blog reports 43,697 subscribers via Bloglines).
Sunday, January 28, 2007
Ask maintains a blacklist of IP addresses
Someone at Ask.com should realize that IP addresses are dynamically given out by ISPs to their clients (potentially) each time they connect - so blacklisting "clients" by IP addresses is not a very smart thing to be doing.
I recently found that the IP address I was now allocated was somehow on Ask's blacklist, and that instead of getting Ask's search engine, all I got was a page saying:
I'm not sure if emailing that address is an automated system to remove the IP address from the blacklist - you can probably spot that the text they ask to be included in the body of the email is simply the hex representation of the client's IP address, (repeated 3 times for some reason). (I've obscured the actual address in the example above, but kept the same format).
Having sent the email, I've certainly not gained instant access to Ask again - I rather think that it will be quicker to get a new IP address from the ISP's DHCP server than to sort out Ask's broken blacklist. If Ask wants to use a blacklist system then they should probably expire IP addresses off the blacklist say a few hours after whatever behaviour it is that triggers the inclusion on the list was last detected. That way, zombie machines that continue to do something bad via Ask will remain blacklisted and hence blocked, but legitimate users who inherit an IP address that was previously used by a zombie are not permanently banned.
I recently found that the IP address I was now allocated was somehow on Ask's blacklist, and that instead of getting Ask's search engine, all I got was a page saying:
Your client does not have permission to access this site.
Please refer to the Ask.com terms of service page.(Ask.com and Syndication).
If you feel that you have received this response in error, please send an email to unauthorized@ask.com. Before sending this email, please refer to our terms of service page, accessible at the url provided above.
Please copy and paste the information below into the body of the email.
fff9480d-2fff9480dfff9480d
I'm not sure if emailing that address is an automated system to remove the IP address from the blacklist - you can probably spot that the text they ask to be included in the body of the email is simply the hex representation of the client's IP address, (repeated 3 times for some reason). (I've obscured the actual address in the example above, but kept the same format).
Having sent the email, I've certainly not gained instant access to Ask again - I rather think that it will be quicker to get a new IP address from the ISP's DHCP server than to sort out Ask's broken blacklist. If Ask wants to use a blacklist system then they should probably expire IP addresses off the blacklist say a few hours after whatever behaviour it is that triggers the inclusion on the list was last detected. That way, zombie machines that continue to do something bad via Ask will remain blacklisted and hence blocked, but legitimate users who inherit an IP address that was previously used by a zombie are not permanently banned.
Sunday, January 14, 2007
Findory rides into the sunset
Searching is for when you know roughly what you are looking for.
An alternative way of finding information is for it to be recommended to you. One of the leaders in recommendation systems is Amazon.
Findory tries to apply some of the ideas Greg Linden learnt when building Amazon's recommendation systems (and what he's learnt since!), but unfortunately Greg has found it difficult to continue to grow Findory - and has just announced that "Findory rides into the sunset". Roughly speaking, this means that development of the site has slowed to a crawl, and that Greg is to concentrate on other things - which for the moment he explains means health and family.
I'm sorry to hear this - Greg's was one of the best blogs in the search area, backed by his ongoing research interests and development of Findory, and whilst I'm sure his interest in personalizing information will continue, without the need to attend to it on a daily basis, I suspect we will be hearing less from Greg in the future. I wish him well, and sorry he couldn't find a way to continue expanding Findory.
An alternative way of finding information is for it to be recommended to you. One of the leaders in recommendation systems is Amazon.
Findory tries to apply some of the ideas Greg Linden learnt when building Amazon's recommendation systems (and what he's learnt since!), but unfortunately Greg has found it difficult to continue to grow Findory - and has just announced that "Findory rides into the sunset". Roughly speaking, this means that development of the site has slowed to a crawl, and that Greg is to concentrate on other things - which for the moment he explains means health and family.
I'm sorry to hear this - Greg's was one of the best blogs in the search area, backed by his ongoing research interests and development of Findory, and whilst I'm sure his interest in personalizing information will continue, without the need to attend to it on a daily basis, I suspect we will be hearing less from Greg in the future. I wish him well, and sorry he couldn't find a way to continue expanding Findory.
Labels: findory, personalization
Thursday, October 05, 2006
Keyhole website still up and active
When Google bought Neven Vision recently they were extremely quick in tearing down all the pages on the websites that Neven Vision had.
I was therefore rather surprised to find that although
http://www.keyhole.com
redirects to Google Earth, if you just add index.html to the url thus
http://www.keyhole.com/index.html
then the old Keyhole site is still there, and apparently fully functioning.
I was therefore rather surprised to find that although
http://www.keyhole.com
redirects to Google Earth, if you just add index.html to the url thus
http://www.keyhole.com/index.html
then the old Keyhole site is still there, and apparently fully functioning.
Labels: google earth, keyhole
Friday, September 29, 2006
OCR problems in Google Books
The accuracy of OCR is vastly increased when the words you read exist in a dictionary. In that case the OCR, when trying to decide between more than one possible interpretation, can use the dictionary to help determine which is the more likely.
As the Google Books project announces its extension to scan European books in Madrid it will have to adjust the dictionary it uses - it's no good scanning Spanish books using the same dictionary as American English books.
However, there are some "English" documents that Google has scanned already where the OCR process has gone very wrong. Consider this page of old printed English, with the "long s" symbol, which looks like a modern "f" character. Looks like the OCR was not told that this dated from 1796, so to look out for long "s" - hence it has identified lots of "fuch" and "fale" rather than "such" and "sale" on the page. A simple dictionary check would have helped here - but only if the process expects "f" and "s" to be confused.
As the Google Books project announces its extension to scan European books in Madrid it will have to adjust the dictionary it uses - it's no good scanning Spanish books using the same dictionary as American English books.
However, there are some "English" documents that Google has scanned already where the OCR process has gone very wrong. Consider this page of old printed English, with the "long s" symbol, which looks like a modern "f" character. Looks like the OCR was not told that this dated from 1796, so to look out for long "s" - hence it has identified lots of "fuch" and "fale" rather than "such" and "sale" on the page. A simple dictionary check would have helped here - but only if the process expects "f" and "s" to be confused.
Wednesday, August 16, 2006
Picasa Web Albums adds email digests
There is a new feature just appeared on the Picasa Web Albums settings page. Between the existing Public Gallery URL, and the Content Controls, there is now a section that allows you to set up an Email Digest.
It looks like Google are defaulting this to a weekly digest, to get the news out to people, and then giving them an option in the email to turn it off. The 4 choices given are no digest, or at a frequency of daily, weekly, or monthly. The description alongside explains:
What is an Email Digest?
When people you have marked as favorites create new albums, upload more photos, or comment on your photos, we send you a summary of these activities to your email address at the interval you specify.
The digest email itself is an HTML formatted email as follows:
The links in the email for changing settings are generic - they just take you to the main Picasa Web page, where you can then log in if needed. This means that you can forward the email to someone else without security issues.
The images do of course link through to the album that has changed. The thumbnails shown are embedded in the email, not just links to the Picasa Web site, so they can be seen when offline, or when external images are turned off (as is the case with most email programs these days).
I didn't have any new comments for this example to notify me about, but the source of the email indicates that they would follow after the list of changed albums.
(This should have appeared on my Documenting Picasa blog, but I'm having a few publishing issues at the moment).
Update:It did eventually make it to the Documenting Picasa Blog.
Labels: picasa
Sunday, August 13, 2006
Most popular tags on flickr
Whilst thinking about Searching Picasa Web Albums I did a bit of investigation into the most popular tags on flickr.
flickr provide a tag cloud, but that is just a visualization of the data, and I couldn't see that they provide the data in a direct form anywhere on the site. However, with the aid of the flickr API I was able to take all the words in the tag cloud, and feed them to the API, to get a count of how many matching tags there were for each word. In addition, I also fed those same words to the "free text" search API call, which looks for the word in more than just the tag field - considereing comments and descriptions for example.
The table below shows 3 columns - the tag I searched for, the number of tags of that word found, and the number of photos found using a free text search. The table is ordered by popularity of tags.
flickr provide a tag cloud, but that is just a visualization of the data, and I couldn't see that they provide the data in a direct form anywhere on the site. However, with the aid of the flickr API I was able to take all the words in the tag cloud, and feed them to the API, to get a count of how many matching tags there were for each word. In addition, I also fed those same words to the "free text" search API call, which looks for the word in more than just the tag field - considereing comments and descriptions for example.
The table below shows 3 columns - the tag I searched for, the number of tags of that word found, and the number of photos found using a free text search. The table is ordered by popularity of tags.
| wedding | 1697455 | 2269471 |
| party | 1475681 | 2350563 |
| family | 1160057 | 1557739 |
| travel | 1113576 | 1275557 |
| friends | 1093997 | 1649955 |
| japan | 1074281 | 1167494 |
| vacation | 1020423 | 1162147 |
| london | 927172 | 1100791 |
| beach | 864728 | 1493985 |
| california | 837012 | 960129 |
| birthday | 829670 | 1371859 |
| trip | 817324 | 1786838 |
| nyc | 774884 | 879161 |
| summer | 732387 | 1122946 |
| nature | 728059 | 816467 |
| italy | 686976 | 776812 |
| france | 669406 | 759215 |
| me | 664111 | 2886782 |
| paris | 662024 | 786996 |
| art | 643563 | 1104574 |
| flowers | 627830 | 1439335 |
| sanfrancisco | 618604 | 620523 |
| europe | 610237 | 708284 |
| china | 604334 | 707039 |
| flower | 596891 | 1439327 |
| newyork | 572049 | 575557 |
| water | 565749 | 973599 |
| people | 552582 | 975146 |
| music | 551493 | 740282 |
| cameraphone | 544369 | 634883 |
| australia | 543923 | 611789 |
| christmas | 537883 | 834827 |
| usa | 531563 | 619628 |
| sky | 527376 | 756538 |
| germany | 526931 | 595724 |
| new | 526549 | 2723772 |
| canada | 512067 | 612492 |
| night | 510466 | 1210035 |
| cat | 504840 | 852699 |
| holiday | 501819 | 703551 |
| park | 498178 | 1820673 |
| bw | 495568 | 514618 |
| dog | 486692 | 859611 |
| food | 478691 | 684164 |
| snow | 478068 | 643189 |
| baby | 463882 | 837637 |
| sunset | 462862 | 752411 |
| city | 446866 | 1294211 |
| chicago | 445639 | 529140 |
| spain | 441158 | 491290 |
| taiwan | 437633 | 461265 |
| july | 435057 | 1033708 |
| blue | 431367 | 862404 |
| tokyo | 427181 | 505507 |
| england | 425920 | 510575 |
| mexico | 406752 | 535804 |
| winter | 403827 | 579849 |
| portrait | 396759 | 644632 |
| green | 394571 | 671410 |
| red | 393719 | 848593 |
| fun | 392680 | 688356 |
| india | 386409 | 446473 |
| architecture | 383526 | 423993 |
| garden | 382303 | 981077 |
| macro | 380663 | 428872 |
| spring | 373441 | 759260 |
| thailand | 368331 | 411644 |
| uk | 362555 | 468979 |
| seattle | 357614 | 421000 |
| festival | 357137 | 658357 |
| concert | 350224 | 471579 |
| canon | 349327 | 580369 |
| house | 347404 | 1236984 |
| berlin | 343694 | 403534 |
| hawaii | 340744 | 384573 |
| street | 339503 | 1036943 |
| lake | 324783 | 937873 |
| zoo | 324442 | 460997 |
| florida | 322534 | 387322 |
| june | 321243 | 729667 |
| may | 316693 | 903231 |
| white | 313703 | 867135 |
| vancouver | 312657 | 383569 |
| kids | 312551 | 635702 |
| tree | 312542 | 1019630 |
| clouds | 307407 | 479446 |
| toronto | 304648 | 359194 |
| barcelona | 293035 | 341533 |
| geotagged | 292631 | 294655 |
| home | 291227 | 778635 |
| sea | 288469 | 546365 |
| day | 287287 | 2136618 |
| texas | 284851 | 360478 |
| scotland | 284615 | 309559 |
| car | 281730 | 788266 |
| light | 281436 | 909673 |
| halloween | 280515 | 362042 |
| camping | 279238 | 654077 |
| church | 273001 | 535265 |
| animals | 270757 | 542946 |
| trees | 270098 | 1019628 |
| washington | 266921 | 470046 |
| river | 266128 | 668903 |
| nikon | 265291 | 393117 |
| april | 262835 | 581807 |
| boston | 261935 | 330032 |
| girl | 258663 | 857847 |
| ireland | 258647 | 305876 |
| graffiti | 257081 | 283560 |
| amsterdam | 256794 | 307304 |
| rock | 256142 | 795001 |
| landscape | 254799 | 343825 |
| blackandwhite | 252987 | 253364 |
| cats | 252971 | 852700 |
| newyorkcity | 250213 | 250430 |
| san | 247739 | 1271943 |
| rome | 246027 | 286082 |
| roadtrip | 245326 | 255484 |
| urban | 244511 | 321849 |
| honeymoon | 241904 | 271611 |
| ocean | 240941 | 359337 |
| dc | 235053 | 363788 |
| newzealand | 232776 | 233573 |
| march | 231491 | 532229 |
| black | 231180 | 708269 |
| museum | 230643 | 562979 |
| york | 230193 | 868773 |
| hiking | 226067 | 398781 |
| island | 224110 | 844505 |
| mountains | 223030 | 759360 |
| yellow | 221515 | 347290 |
| sydney | 221447 | 278644 |
| sun | 218685 | 503464 |
| hongkong | 216193 | 222298 |
| show | 216155 | 850535 |
| graduation | 214681 | 306454 |
| color | 213967 | 579521 |
| film | 210509 | 351853 |
| mountain | 210046 | 759360 |
| animal | 208416 | 542946 |
| losangeles | 207932 | 208309 |
| school | 205891 | 550486 |
| moblog | 205613 | 210954 |
| photo | 205047 | 2308835 |
| dogs | 203610 | 859613 |
