This is a very interesting proposal! Because this is very ambitious as you mention, I think that having some contingency plans in place would be a good idea. One possible version might not necessarily involve computer vision -- for example, the algorithm could simply take as inputs descriptions of the user's clothing items ("red shirt," "brown pants," etc.) and return recommendations based on these text inputs. This would allow you to have a usable product even if the computer vision component turns out to be more challenging than expected.
However, I do think that the computer vision angle is likely to be possible, so if you would like to focus there then I am all for it. I look forward to seeing what you can produce!
This is a very interesting proposal! Because this is very ambitious as you mention, I think that having some contingency plans in place would be a good idea. One possible version might not necessarily involve computer vision -- for example, the algorithm could simply take as inputs descriptions of the user's clothing items ("red shirt," "brown pants," etc.) and return recommendations based on these text inputs. This would allow you to have a usable product even if the computer vision component turns out to be more challenging than expected.
However, I do think that the computer vision angle is likely to be possible, so if you would like to focus there then I am all for it. I look forward to seeing what you can produce!