mvoelk/ssd_detectors

Understanding the flow

Opened this issue · 5 comments

Lakshaysharma048 commented

Hi mvoelk,
I recently saw your code and unable to understand from where to start.

Unable to find main function.

Can you please help.

mvoelk commented

If you are familliar with Jupyter Notebook, the SSD related notebooks SSD_predict.ipynb, SSD_train.ipynb and SSD_evaluation.ipynb should be a good starting point. I will also update the code in the next few days...

Lakshaysharma048 commented

Hi Mvoelk,

Thank you.
Is there anything that is missing in SSD text detection part?

mvoelk commented

SSD is an generic object detector. For the text detection part, see TB_*.ipynb, SL_*.ipynb and TBPP_*.ipynb. For more details, please read the README.md, the related papers and the thesis. The current code should run as it is.

Lakshaysharma048 commented

Hi Mvoelk,

I was able to understand the TextBox code but at the prediction time, it is giving the class label as "text". (Text and Background) is what we have set as class labels.

I want to pass the output to CRNN function which in turn will check it's lexicon to give the exact word name back to the textbox code. is there a way to connect both of them.

mvoelk commented

In SL_end2end_predict.ipynb, the text instances are cropped and passed to CRNN. The CRNN implementation is lexicon free.