Generate imdb file from annotations file

[nascimentocrafael]

Hi,

I'd like to train my own object detector with fast-rcnn, but I can't figure out how to generate the imdb file with roi annotations for training. Could you please show me some pointers on how to do that? How do I create a imdb file from a txt file like this image_name.jpg x1 y1 x2 y2 class My goal is to have a imdb file to use in train.py

Thank you in advance.

[ssakhavi]

I would refer you to the issue #11 , but I'm guessing that it isn't the answer to your question.

Actually, I have the same problem as above @rbgirshick. Maybe I can rephrase the question in a more general way:

In what format and where should the data be written in order to be used in the Python Layer?

[acpn]

Anybody managed? Same problem here.

[NightFury13]

Still looking for this answer. Any updates @acpn @ssakhavi @nascimentocrafael ? It's quite weird that there's a lot of post discussing the 'changes' they made to the original script to get things running for their use-case but no post discussing what is the 'format' expected by these scripts. Reading through code to figure out the format is not something everyone should be expected to do. Will post the solution here if I figure this out :slightly_frowning_face: