The Roads Less Travelled …

Couple of days back, I finally recieved my long awaited invitation to try out Riya. And after the long wait, I had to get two invitations; one from the official Riya team and another from a friend who has just joined their team here in Bangalore (in case you were wondering, having friends in right places still helps ). So today evening I started the uploader install and immediately afterwards started uploading my photos.

The first experience with Riya, I have to admit was not great. On what was not great I will go into detail a little later but there were moments when I was literally laughing my guts out. Especially when Riya grouped together a friends’ face with that of Aishwarya Rai’s from a poster present in another photograph.

My entire collection would be like 900 pictures or so and each person would on an average feature 5-6 times. I am mentioning this because I am not sure how large a training set is needed for Riya to function optimally and hence the comments following these lines would have to be taken in this context.

Uploader

• Very slow install. Is it a web install like the MSN and Yahoo messengers? By comparision to these also, it takes a lot of time. And what does Riya take 30 MB for? Not the uploader only, for sure?
• No choice on where to install the uploader. I know people in Riya do not have evil designs on my poor laptop but I remain a e-hypochondriac.
• I stopped and restarted the upload a couple of times. Both the times the existing state of the upload before my stopping it was lost; not on the server but on the uploader UI itself. After a couple of such timeouts (mainly for compiling my work) I lost track of how many pictures were done or how many were left.
• The upload process itself was pretty fast considering they do the detection of faces and text at the same time.
• Sometimes the face recognized the picture as shown in the uploader UI is an extraneous face not present in the picture. I am pretty sure of this having encountered it more than a couple of times.
• A suggestion: Can I tag my pictures from the uploader itself??
• Another one: Can it scan my HDD and take all my pics? After all they are private to me only, right?

Face/Text Detection

• I would say Riya does a really good job in face detection. In 900 odd pictures I had < 10 instances of marking a recognized entity as “Not a Face”.
• I did not have too much of text in my photographs, so can’t really comment on this feature but any text I had in some was not recognized.

Face Recognition, Tagging and General workflow

• When can start tagging and playing with my pictures? My guess was as soon as the first one gets into the Riya server; but I have a hunch now that it might be better to allow the uploading to finish before starting on the play. Apart from it taking time and CPU resources, does the suggestion of letting the uploading happen overnight obliquely avoid this problem also?? [Too bad I am an owl]
• After spending around a dozen rounds of manual training (single and bulk), I am not sure if I could not have tagged my pictures with the necessary meta data for effective searching later, in any other application. In short, recognition did not work too well for me. I can understand cases where I have only 5 or 6 cases of the face, but even with say 15 instances it asks me again to identify the, not too distinct, next one. And worse I am pretty sure I tagged some faces more than once.
• Auto complete for tagging is too buggy. Sometimes a dropdown appeared and sometimes it did not.
• When I see only one picture and tag one of the faces there, Riya immediately informs the user on the size of the training set. Somehow for most pictures the training set size was shown as one to me; though I am pretty sure of having tagged more than one picture with the tag in question. Is a training set any different from the faces one has manually tagged?
• I lost the tags created in one manual training session once; which was again not a very great experience.
• The search in “My Photos” did not work for most of the tags that I had created. Very irritating since Riya is all about search.
• The RSS feed of the search looks like either a RSS 1.0 feed or the RDF standard. Why not RSS 2.0 or Atom??

I have not had the time to yet experiment with their sharing and the public search feature but will do that soon.

Building a software product is a gargantuan task but people in Riya are trying to build something for which there is no precursor and hence their task is doubly hard. Not only does this product involve leaps in technology but also new UX models. The welcome email as well as the site says very clearly that the intelligence Riya has now is akin to a two year, and that we should not over expect. I can realize now, why the caveat is there. I just hope that Riya does not fall into what Joel’s Marimba phenomenon.

Technorati Tags: riyasobetaithurts

Sometime back I read Women by Bukowski and had decided that unless I was really short of books, I would not pick one of his again. But I have and do not grudge my choice at all.

Factotum like Women centres around Henry Chinaski; however the key difference is that while in Women Chinaski was rather well established in Factotum he is still the roving wanderer, an undiscovered artiste, doing odd jobs, getting drunk, being fired and all the while trying to write. And while he is still repetitive (the entire book can be read as Chinaski’s job history), it did not get boring as Women tended to become at times. The primary reason, I guess, was that shifting from place to place, doing odd jobs - from being a janitor to shipping clerk - Chinaski’s account here has a dynamism that, I felt, Women lacked.

Both of the books I have read by Bukowski have dealt with journeys. In Women it was the journey of relationships and in Factotum it is search to find the career that the protagonist wishes for. Neither of the works end with a goal reached (not even one in sight), however they do leave one with the feeling that the end is not too far ahead. I guess it is upto the reader to fathom if the end is real or merely a mirage.

At levels Bukowski’s work can be seen as the Catch 22 situation that unestablished artists face. To create art you have to live life your way and see the world as you want to see it. However good writing does not come from an empty stomach, and Bukowski does try a lot to dispel the myth of the starving genius. And to ensure that you hold a steady job and write well, you really can’t go on drinking binges and tell your boss what you think of him. Nowhere is this paradox as evident as in Factotum.

Underplayed satire coupled with the incisive, unabashed and telling observations about the society form the bulwark on which the book stands. In one case, where Chinaski’s job is to pack brake shoes into different cartons (graded on their quality), he says that he does not really know which are which; to which he gets the curt reply from his boss that all are the same. The incident brings out very well the major strengths of Bokowski’s writings.

Personally, this is after a considerably long time, maybe close to a month, that I finished a work of fiction. I was thinking that maybe I had outgrown fiction. Thankfully, there is a always a good book around the corner to pull you back in. And after reading this, I am thinking of getting hold of some more works by Charles Bukowski. Top of my list would probably be Post Office.

TechCrunch published yesterday screenshots of the upcoming Google’s Ajax calendar application, CL2, here. Rumours have been around long that Google is developing such an application and from the screenshots it looks that the smoke was not without any fire. And today Om Malik has published rumours heard in the corridors of money that Google is set to acquire Writely. With CL2, Gmail and Writely, Google office on the web might be a reality sooner than many expected.

Apart from its tight integration with GMail, the ability of CL2 to discover new events is an indication of the differentiating features of this application vis-a-vis the numerous other such calendar applications. And this seems to be the first time Google has jumped into the sharing space, a move that I believe has far greater ramifications than just being a killer calendar feature. Apart from the capability to search and discover events and to deliver notifications via a gamut of media (mail, SMS) will really set this product apart.

The slides at the Google’s analyst day discussing GDrive coupled with the Writely rumours gives me a feeling that the ePIc video predicting a Google/Amazon grid around 2014 is well on its course to reality. Today Google seems to be the only player capable of delivering on this vision; by leveraging and upgrading the massively parrallel system they have built. While the applications themselves can be copied easily, delivering the scalability and reliability of Google will take considerable time to achieve; though many including Microsoft have their hopes.

What does this bode for the numerous Web 2.0 startup each tackling at varying levels, the problem of delivering user content and services on the Web.?They will either be bought over or will die, one can assume. By ensuring interoperability of data between its various applications (Gmail, CL2, Google Base, GDrive) Google and other big players will have an advantage that cannot be replicated very easily. The only hope for the smaller players lies in the adoption of global standards in terms of how content is published and delivered so that their applications achieve smooth interoperability with others. However even with global standards the tricky problem of how data is modelled by each of these applications will probably still remain limiting how useful synergies between small independent players can be.

The prospect of a Google grid is at once frightening and exhilarating. That we are on the path of freedom from desktops seems evident now. The question remains on what exactly is the utopian/dystopian end of this path? Are we seeing the rise of the biggest brother of them all?

If one has to take a message out of the Oscars this year, it would be “Go watch movies in the theatre and not DVDs and definitely not pirated ones.”; not surprizing since the theatre revenues fell enough in 2005 to warrant the Academy’s august attention. Those hunting for reasons can find some here.

I see the problem as a cyclical chain. Reduction in the number of people going to the theatres greatly diminishes the experience of watching a movie there, which in turn keeps away more viewers. As the preseident of the Academy said, watching a movie in a theatre is akin to sharing your emotions with hundreds of strangers; and the opportunities for such platforms are not many in todays world. What makes the Govinda movies funny is not only his over-the-top-comic actions but also the comments, hoots, jeers and shouts from the general public. What transforms a Lagaan into a rivetting tale is not the script but the wishing of the crowd that Bhuvan hits the last ball for a six, as if it was not a movie but a live cricket match going on. Theatre adds a different dimension to the experience of watching a movie.

Sadly, such experience is proving to be a costly affair these days.

Match Point, has been called alternatively Woody Allen’s return to form or a break (not a great one at that) from his unique style of filmmaking. One thing is certain, Match Point from the top does not look like a Woody Allen movie but is nonetheless a cracking movie. And once we dig a little we find the same themes that Allen has always talked about in his previous flicks.

Allen deserts his muse New York here; moving over to her distant cousin London across the Atlantic. There are none of the standard Allenisms in terms of the overpowering satire or biting sarcasm in the movie; or no judgements on a society malformed. In fact, it seems that Allen is finally trying to tell *only* a story. And he does that well. And for this reason I would say Match Point is a good movie.

However as much as he has sacrificed in terms of style and practices, he has embraced his core themes that much more closely. Allen has grown old; he seems to have lost the hope that tearing apart the existing structures and conventions with his mordant wit is possible. He presents the characters so enmeshed in this societal structure that they themselves murder any hope of an escape from it. The same themes as before, merely a new way of looking at them.

An excellent article on InternetWeek comparing four of the most popular browsers; IE 7, Firefox, Opera and Maxthon is here.

Having used all of these and some more, I personally stick to Firefox because it allows me to work with my information from various sources, Flickr, del.icio.us, WordPress in a seamless manner.