search

yanconglin / Deep-Hough-Transform-Line-Priors

Official implementation for Deep-Hough-Transform-Line-Priors (ECCV 2020)
MIT License
161 stars 31 forks source link

Format of training label #6

Closed omrastogi closed 2 years ago

omrastogi commented 3 years ago

Hi Yancong, Its needless to say what a incredible job you are doing.

I am trying to write a data loader to train the model. Can you give some idea about format of label data. How can I train my custom dataset? (The datasets for this project is unfortunately not available)

I would like to know, essential information needed to train the model. Specifically I don't understand the following:-

Screenshot from 2021-09-23 13-16-36

See, if you can help me. Thank you in advance.

omrastogi commented 3 years ago

Here is what I already understand: The structure of the dataset is as follows: data/wireframe --> train(folder) ----> label for train image in '.npz' format --> valid(folder) ----> label for validation image in '.npz' format --> images with same name as their label counterparts in train/valid folder, having an extension of ".png"

yanconglin commented 3 years ago

hi, Omrastogi,

Thank you for the interest. Please check this: https://github.com/yanconglin/Deep-Hough-Transform-Line-Priors/blob/bbe6b8f533e81070cb2880600901a38eeb58cc9d/ht-lcnn/dataset/wireframe.py You will find all the details about the data format. The best practice would be: (1) download the dataset, and run the script (mentioned above) by yourself.

Yancong

yanconglin commented 2 years ago

Feel free to re-open if you have further questions.