Had an interesting look into detecting changes in two images yesterday. I'd known about the algorithms that are used to detect motion in video by analysing the frames. So essentially I decided to see if I could compare two images this way, treating both of them as frames in a movie. Interestingly enough it worked, taking multiple screenshots of my desktop I was able to detect the 'changes' between them moreover I was able to detect the exact regions that the change occurred. This allows me to overlay the second image with an indication of which 'objects' are new or old. These are Neely represented by a rectangle around the region in question. Right now, I've not identified a region as being 'new' or 'old' at this point.
This technology may have wider teaching consequences than just showing the difference between two images: I see the ability to compare two screenshots of two versions of internet explorer browsers rendering the same page of a website being able to automatically show CSS(alignment /layout/presentation) differences between the two along with other apparent changes in rendering across two browsers.
The other interesting one for me is slightly geekier: the thought that one could detect people, objects etc in motion frames. This interests me particularly as I've been wondering about how one would implement software to function like Jarvis - Iron man's Robot, doh!
So about that, I was thinking a Jarvis implementation should really be noddles around obtaining information and then linking information together so that one almost stores a bunch of the same or related information together and then link it together somehow in a smart way - much like how I imagine the human brain might deal with incoming information.
Information needs to be common in some respects. Information from sight may include common format which wraps colour, depth and dimension around the 'what' is being seen.
Information from sound might include pitch & tone, object reference that's producing the sound and what that sound means - the 'what' in sound communication. Other information might come from the Internet, manual submission etc but still, one would need to wrap obtained information with as much auxiliary information that relates to and provides context for this information to make sense and more importantly be broken down and analysed when trying to make connections between two pieces of such information.
I do think we have fundamental rules and categories however dynamic that store information we get into these categories which then like to the underlying information. I can't be sure but it may lead to a way to automatically obtain, store and relate information without manual intervention or external data entry by say me - IE no aid in storing information and making links.
Neural technology, fuzzy logic and image recognition like described earlier might, if implemented as the best software program possible may make Jarvis possible. What's more it might be possible to create yourself a second brain to manage the vast amounts of information out there, on the Internet and information you present to reasoning and deducting so as to give you results about information quicker.
It won't be fool proof but if it works, whatever you get - if it helps without hindering and is practical then job done.
This technology may have wider teaching consequences than just showing the difference between two images: I see the ability to compare two screenshots of two versions of internet explorer browsers rendering the same page of a website being able to automatically show CSS(alignment /layout/presentation) differences between the two along with other apparent changes in rendering across two browsers.
The other interesting one for me is slightly geekier: the thought that one could detect people, objects etc in motion frames. This interests me particularly as I've been wondering about how one would implement software to function like Jarvis - Iron man's Robot, doh!
So about that, I was thinking a Jarvis implementation should really be noddles around obtaining information and then linking information together so that one almost stores a bunch of the same or related information together and then link it together somehow in a smart way - much like how I imagine the human brain might deal with incoming information.
Information needs to be common in some respects. Information from sight may include common format which wraps colour, depth and dimension around the 'what' is being seen.
Information from sound might include pitch & tone, object reference that's producing the sound and what that sound means - the 'what' in sound communication. Other information might come from the Internet, manual submission etc but still, one would need to wrap obtained information with as much auxiliary information that relates to and provides context for this information to make sense and more importantly be broken down and analysed when trying to make connections between two pieces of such information.
I do think we have fundamental rules and categories however dynamic that store information we get into these categories which then like to the underlying information. I can't be sure but it may lead to a way to automatically obtain, store and relate information without manual intervention or external data entry by say me - IE no aid in storing information and making links.
Neural technology, fuzzy logic and image recognition like described earlier might, if implemented as the best software program possible may make Jarvis possible. What's more it might be possible to create yourself a second brain to manage the vast amounts of information out there, on the Internet and information you present to reasoning and deducting so as to give you results about information quicker.
It won't be fool proof but if it works, whatever you get - if it helps without hindering and is practical then job done.