Over the past few weeks, I have taken a liking to Visual Information Retrieval and have read few papers on the concept, algorithms etc behind this. Also I've been recommended few good libraries on Python which support CBIR. Those interested can read on the same on Wikipedia. Apparently there are many proprietary algorithms used in this field, however its still not clear where and how do these patents are enforceable?
I also came across a good Java library called LIRE which is open source and provides most of the features that one would require to implement a good CBIR system. Apparently the LIRE library also provides open source implementations of SIFT and SURF algorithms, so its still not clear how a patent is enforceable when there exists an open source implementation for the same!
However this isn't a technical blog, nor does it document my experiences in using LIRE for CBIR tasks(which warrants a separate blog by itself, IF and WHEN I complete my research)! I document here, the reason behind my liking to the concept. Well, many of you might consider this blog boastful (but again in a blog titled "Memoirs", what else did you expect ;-) ??). For long, I have prided on my ability to recognize faces and other objects that I've registered in my memory. Many of my friends would agree that I have surprised them(and the person in question) quite so often by recalling a face or a location so vividly when they could hardly recollect or recognize the same(even-though all of us would have seen the same thing/person at the same time)!
Mind you, by no means to I claim to have eidetic memory!! That would be absurd, I only claim that I seem have the ability to register and recall few images better than the rest. I am not sure if this is an inherited trait or an acquired trait. However, I do know that my brother too has similar ability and he had found a better uses of this ability to memorize hard to pronounce Sanskrit slokas in Class IV (Chayagrahini Rakshasi) by just registering a mental image of the page containing the unpronounceable words :-D.
Like most human beings I have a pretty good radar(should I call it Murphy's Radar?). This radar of mine has been able to predict to an acceptable accuracy on the occurrence of an event. However off-late, my intrinsic CBIR library hasn't functioned optimally during such instances, so my strike rate of producing accurate results has gotten poor! In the computer world, if we find the performance of a particular library is not up to the threshold limit, you can always replace it with a better library, But how do we do this in the real world??
I also came across a good Java library called LIRE which is open source and provides most of the features that one would require to implement a good CBIR system. Apparently the LIRE library also provides open source implementations of SIFT and SURF algorithms, so its still not clear how a patent is enforceable when there exists an open source implementation for the same!
However this isn't a technical blog, nor does it document my experiences in using LIRE for CBIR tasks(which warrants a separate blog by itself, IF and WHEN I complete my research)! I document here, the reason behind my liking to the concept. Well, many of you might consider this blog boastful (but again in a blog titled "Memoirs", what else did you expect ;-) ??). For long, I have prided on my ability to recognize faces and other objects that I've registered in my memory. Many of my friends would agree that I have surprised them(and the person in question) quite so often by recalling a face or a location so vividly when they could hardly recollect or recognize the same(even-though all of us would have seen the same thing/person at the same time)!
Mind you, by no means to I claim to have eidetic memory!! That would be absurd, I only claim that I seem have the ability to register and recall few images better than the rest. I am not sure if this is an inherited trait or an acquired trait. However, I do know that my brother too has similar ability and he had found a better uses of this ability to memorize hard to pronounce Sanskrit slokas in Class IV (Chayagrahini Rakshasi) by just registering a mental image of the page containing the unpronounceable words :-D.
Like most human beings I have a pretty good radar(should I call it Murphy's Radar?). This radar of mine has been able to predict to an acceptable accuracy on the occurrence of an event. However off-late, my intrinsic CBIR library hasn't functioned optimally during such instances, so my strike rate of producing accurate results has gotten poor! In the computer world, if we find the performance of a particular library is not up to the threshold limit, you can always replace it with a better library, But how do we do this in the real world??
Secondly how do we identify the cause of the malfunctioning algorithm? Has age got anything to do with it? Or I need a better index? Will I be in a position to re-achieve the optimum results out of my intrinsic CBIR library that I had come to expect from it all these years? Will I manage to hold on to one of the last few things that I have prided upon for few more years?? I guess only time will tell............
