SUYEgit / Surgery-Robot-Detection-Segmentation

Object detection and segmentation for a surgery robot using Mask-RCNN on Python 3, Keras, and TensorFlow..
Other
181 stars 84 forks source link

VIA dataset #4

Closed david-schweitzer closed 6 years ago

david-schweitzer commented 6 years ago

Hello, could you provide a snippet of your VIA annotations? I have a script that creates annotations on the fly based on what's in the data/surgery/... folders, and it's mostly matched up with what VIA does individually, but a few lines I'm not certain about.

For example, in surgery.py on line 165, you have

"name" is the attributes name decided when labeling, etc. 'region_attributes': {name:'a'}

class_names = info["names"]

But from the comment, there is no "names" key in "region_attributes." And these are not necessarily universal or required labeling means in VIA since VIA has apparently a few ways of identifying what regions are.

SUYEgit commented 6 years ago

Thanks for your comment. Yeah you are correct. I really appreciate your reviewing. That line of code is forgotten to be moved to before line 172. 'names' are defined in function add_image. 'name' is the attribute name which is self-defined when labeling. I will update it soon! Thank you!

david-schweitzer commented 6 years ago

This issue can be closed since everything can be found here (for others): #3