mvoelk/ssd_detectors

[Question] Can you please upload the pickled dataset that you used for training SegLink?

Closed this issue · 5 comments

sniper0110 commented

Hello,

Can you please upload the pickled dataset that you used for training SegLink?

It would be great if we can just run the code first and then try to understand the pipeline. I am asking this because I am finding it difficult to understand how the data is prepared for training SegLink. I have trained object detectors before but I think I am missing a step when it comes to training text detectors. So checking your data and playing with it would definitely help me understand better the pipeline.

Thanks in advance.

mvoelk commented

You do not need the pickled data. Simply replace

with open('gt_util_synthtext_seglink.pkl', 'rb') as f:
    gt_util = pickle.load(f)

with

 gt_util = GTUtility('data/SynthText/', polygon=True)

The gt_util_synthtext_seglink.pkl is only to speed up the parsing of the dataset. It is serialized in datasets.ipynb and does not contain the image data itself, only filenames, bounding boxes and so on. See also #1 ...

sniper0110 commented

Thanks for the quick reply. Did you get the SynthText dataset from here http://www.robots.ox.ac.uk/~vgg/data/scenetext/ ?

mvoelk commented

Yes!

sniper0110 commented

Alright thanks!

anindya7 commented

Could someone who has generated the .pkl file PR it? I'm on a limited bandwidth network and would like to run the end2end notebook.