search

imatge-upc / retrieval-2016-icmr

Bags of Local Convolutional Features for Scalable Instance Search (ICMR 2016)
https://imatge-upc.github.io/retrieval-2016-icmr/
111 stars 36 forks source link

how to build? #1

Open hlzz opened 7 years ago

hlzz commented 7 years ago

How to build this project? I can find the requirements.txt file.

evamohe commented 7 years ago

@hlzz, I have just added the requirements.txt missing.

hlzz commented 7 years ago

Is your faster rcnn code a requirement or dependency for this repo?

evamohe commented 7 years ago

It is not a requirement for this repo

willard-yuan commented 7 years ago

@evamohe Hi eva, in settings.json, there is two files, i.e. list_oxford_horitzontal.txt and list_oxford_vertical.txt. Would you like to tell me how can I get these two files?

willard-yuan commented 7 years ago

@evamohe It seems that images in the list_oxford_horitzontal.txt are the images which the width is large than the hight, and the list_oxford_vertical.txt are the images which the hight is large than the width. Did I understand it exactly?

Paul0629 commented 6 years ago

@evamohe Dear Eva, I'm reading your paper and code. I have the same questions with willard-yuan about the list_oxford_horitzontal.txt and list_oxford_vertical.txt. Could you help us?

evamohe commented 6 years ago

Hi @Paul0629,

"list_oxford_vertical.txt" and "list_oxford_horitzontal.txt" are lists with the absolute path with portrait/landmark aspect ratio images within the dataset.

You can take a look and slightly modify for your dataset the preprocessing script (https://github.com/imatge-upc/retrieval-2016-icmr/blob/master/oxford105k/preprocess.py) in in the Oxford105 folder to create these files.

Paul0629 commented 6 years ago

Really appreciate your help. Many many thanks.@evamohe