Vamsidhar Reddy Gaddam
Smart Machines: Can they help us see better?
University of Oslo, Department of Informatics
From early in the education, a programmable calculator seemed extremely helpful in accompanying me for practical tasks where I could focus on higher level problem solving. I always appreciated using machines for managing practical tasks – at times there are softwares available readily and at times I had to write small scripts and hacks to accomplish them. My utmost fascination came when I interfaced a few sensors with a processor and got it to perform some ‘intelligent’ tasks being aware of its surroundings. Ever since, I got stuck to the collaboration of cameras and other sensors with machines.
My PhD project
When it comes to certain fields, a lot of work is still performed manually.
Taking medical examinations for instance, we still have doctors running through the entire footage to be able to spot a tumor or an anomoly – the training for which happens by examples. Taking football broadcast, the cameraman still moves the camera manually- even though the movement dynamics are pretty much the same every game. We are developing a framework to provide higher level abstraction to such systems and learn from the knowledge of the experts. Selection of football as a field to experiment is merely because of the general interest and expertise an average citizen can have on the game – which in turn provides us input to our system in chosing different algorithms by user studies. We are at a stage where we proved this as a concept and are extending it to other fields.
My career dream
Humans developed a complex visual analysis system.
With supervised training, one can expect similar results from machines. We are already able to produce cameras that can capture at much higher rate and higher quality than our eyes can see. But the magic in human visual system is the brain and today we can atleast train systems to help us perform certain critical visual tasks better and in future, they can even be able to perform them completely autonomous.
This will revolutionize the ease of diagnosis in remote places, where coke and a cell phone have penetrated but not medical professionals.