HAHA. As a PhD student, I agree with most of your sentiments. The PhD application process is nothing, try applying for fellowships. Go to a prestigious institution, pull allnighters, to get a subpar gpa, and have no time for volunteering activities. These are the criteria you need to satisfy to get something like NSF GRP (Broader Impact is 50% of the application). You might have a better shot at a lesser prestigious university with a not-so-great engineering program and being a super smart person, killing it. Then, apply for fellowship --> you will get it. It is all a rubric.
Now on to the game of academia. It is one that is based on reputation, papers, and not money. No one cares for efficiency either and no one is building crap. All they are doing is trying out different "solutions," which may or may not make any sense. For instance, "hey guys, let's apply game theory to problem X, it will be a good/novel experiment and we can get a paper out of it." Waste of tax dollars. The logical answer that solves problem X might have already been solved and can be solved using ordinary methods, but academia doesn't want that. Research labs have business models, keep proposing new complexity on complexities, very unnecessary. Overcrowded fields like computer vision reek of this.
Don't even get me started on funding and fellowships haha. I think trying to get funding in math must be the worst. Every formula has to boil down to "security this" or "missile navigation that."
They talk about your attitude when you were talking to them. Are they convinced by your eye contact and shoes? If you don't take care of your appearance, you won't meticulously consider the niceties required to face customers and leading on gently around your products.
Choosing 2 of 3 is true for early-stage startups. But once things take off, then you can get 3 out of 3 as you are able to delegate some of the work out to your employees. Then, you can achieve startup, health, significant other/family.
Yes! I hate it when startups think they are so cool by using Facebook/Social logins. Sane users have one good email account for work/important stuff and one useless email account for experimental services such as this. Why would anyone want to give out all their Fb info to random companies.
It also makes you one dimensional because everyone is like you. This can be bad for your own personal development and coming across ideas in different areas of industry/economy. No wonder you see "herd" startups in SF a lot of times. There is too much noise in SF.
Regarding the segmentation phase, we were not satisfied with any existing algorithms because we wanted the tool to be as real-time as possible (for up to 4 megapixel images), so we rolled our own segmentation algorithm that has some similarities to GrabCuts (not GraphCuts).
But the binary segmentation is really only a small part of the technical challenge. You also need to figure out how to anti-alias the boundary, and remove the background halo. Simply feathering doesn't do the job unless you also erode the foreground first, which we wanted to avoid. So we developed a custom sub-pixel background-removal and anti-aliasing algorithm.
I think you guys did a great job. This is a very tough problem (speaking from experience) particularly depending on the complexity of the image. One problem we had to solve along the way for our startup is similar, but focused only on fashion related images. Quite tricky particularly if you want something automatic. Anyway cheers! :)
This is a case of coherence vs coupling, which is lost by OO developers. Everything is so tightly coupled that more spaghetti code is written up to not break the coupling, making the code more incoherent overall.
