Assembling a 750 Jigsaw

On 19th June, we started assembling a 750 pieces puzzle. And then almost everyday, the regular exercise once I am back from office would be to work on it. We'd spend almost 15-20-30 minutes everyday just before dinner. And exactly five weeks later, we placed the final piece in the puzzle, making the picture of a Moose Deer complete.

The strategy first was to divide the pieces into as many groups as possible; and that is done by similarity of colours or patterns. So we took around 15 envelopes, and then created as many heaps of similar looking pieces, and each heap went into one. The idea being members of each envelope would be in close proximity of each other. This is particularly important that all similar tiles are together. One tile in wrong heap will cause a long search. This was not 100% accurate, though. There would be several disjoint areas which might have very similar colours; especially almost black tiles; and they are going to mislead. However, we can deal with such special case later. One of the biggest challenge in the whole exercise is not to lose a single piece. Since you work part-time of it, it takes weeks to finish the project, and one single missing tile will spoil all the fun. Moreover, you have to save the work for the next day, which essentially means you have to reserve an undisturbed large area to store the half-finished puzzle at home (and if you can't then ensure you can easily move the work in progress).

The photograph printed on the box was of very limited help. It was very small compared to the size of the actual puzzle; so looking at individual tiles it was very hard to guess which area they are going to go. Still then, the fun was in guessing and finally finding it right.

It'll be painful to disassemble it.

Geotagging Photographs

Technology exists. But many times they aren't integrated, or aren't cheap to use. Several times two independent technologies spring up to serve two completely different purposes; and then someone comes and stitches them together to open up completely new possibilities. Take photography and geolocation as example. Sometimes it is important (or fun) to know where exactly the photograph is taken. In nature photography, for example, you shoot a rare species and try to publish a record. The exact location will provide a much better details on the subject's habitat, or help someone to follow your trail if the want to do further research. Or maybe just as a tour guide. You hiked around, and on the way kept shooting. And the photos are placed in the map exactly where they were taken. Flickr, Picasa had such support since long; if the image had location information in their header, they'd show the image on a map. Even manually you could place photos on their maps and they'll update the meta data.

Obviously updating the location manually is very painful process. All your 100 photos may be taken at different locations (nearby, but different). Updating in Flickr/Picasa just updates their own copy. Plus, it may be difficult to pinpoint the exact location of each photograph in their maps. So, we need to devise a method where the photos will be automatically updated with location information.

A camera with integrated GPS solves everything. Theoretically, though. The cameras with built-in GPS is neither good as a camera nor the GPS works (first hand feedback of a friend who's using one). I already have invested so much in good camera gear that a downgrade (the camera model with gps is way inferior) is out of question. And I have multiple of cameras with me during a trip so obviously I can't change all of them. And the GPS module wouldn't work when needed.

So what I need is an external GPS module. It need not have maps as I am not going to use it as guidance. All I want it to keep logging my location during my photography sessions; and later I can reconcile the log with photo's time. I was also looking at a watch which can monitor my heart-beat during my work-outs, and finally found a deal on Garmin Forerunner 305 at amazon. (They made the watch to help athletes but see the purpose it is serving here). It can log enough to cover a few safari trips, and battery lasts a whole working day. The only missing part in this is the direction the camera is facing when shooting (which an integrated box could do), but I guess I'll live with that deficiency.

This is it, so I synchronized the clocks of all cameras with the GPS watch. Once back from the trek/trip, it is time to update the location information. Currently I am using geotag to stitch the information together. First I export the GPS track into a gpx file from the watch. Then geotag loads the images, the track and places the images on the track as per the time-stamp. It can even interpolate if the photo doesn't have a matching time (rarely as the gpx track will generally record the position every second). And finally update the photos' exif tags with latitude, longitude and altitude information (both RAW and JPG). Now we can upload the photo anywhere, and the meta information would go with it.

There's geotag (right) in action, along with the placement of the photo in google map (top-left) with gps information available from Garmin training center (left-bottom).

Filtering the RJ bakwaas

As the quality of RJ chatters in FM radio is deteriorating every day, the usual "cut craps from life" thoughts come pouring in. When most of the songs they play are pretty good, except those particular ones which are known to be exclusively for FM channels, like item numbers but except visual stimulation they depend on rythmic drums; the talks that connects the audience to the programs are obviously failing to meet even minimal expectations. So just like AD blockers, we need silence when the Radio Jockeys deliver their over enthusiastic speech. Since song titles in FM radio is still a distant reality, we have to do some screen scraping ourselves. So the problem boils down to how to detect a song easily (remember the car stereo has limited processing power and memory). So the biggest difference between talk and song are background music. The human voice will be limited to 2 to 8kHZ frequency range, when the music instruments would produce frequencies in 10 to 20kHz in addition. So easiest rule you can apply is "unless there is reasonable amplitude above 8kHz, mute". In a wide variety of cases, it ought to solve the problem. You can hear the ringing tone from Ghanta Singh, but enjoy golden silence when he bores his victims.

Obviously such oversimplification won't cut the ads. And it'll silence the traffic update announcement too.

Still then, the biggest problem is they won't give me the source code or SDK of my car audio.