Relja / netvlad

NetVLAD: CNN architecture for weakly supervised place recognition
MIT License
534 stars 122 forks source link

yael_dummy error #32

Closed WillTeng closed 4 years ago

WillTeng commented 4 years ago

hello,when i run tiny tokyo time machine datasets,it show assert error in yael_nn file at line 4 and my plateform is window

Relja commented 4 years ago

This function searches for k nearest neighbour in a database of vectors. The assert states that the database size should be larger than k (otherwise it doesn't make sense). So I'm guessing something went wrong with feature extraction and you have an empty database. Do you have the actual Tokyo Time Machine images (if not the feature extraction will return empty)? If you need the place recognition datasets, please email me at relja@relja.fino.

WillTeng commented 4 years ago

I use Tiny subset of Tokyo Time Machine that contains 360 images,is that not work?

Relja commented 4 years ago

So you have the actual images? If yes, then it's strange this gives an error. Are you sure everything else worked fine without errors (i.e. you installed matconvnet and it works) ?

WillTeng commented 4 years ago

I just compile matconvnet again and test it,it works successfully

Relja commented 4 years ago

Good

WillTeng commented 4 years ago

I add an equal sign to the assert,it work normally.Is this data problem?or sometimes k can equal database size?

Relja commented 4 years ago

The function returns k nearest neighbours out of the database. You are forcing k to be the same as the database size.

WillTeng commented 4 years ago

not same,it is smaller or equal