extra labels being detected during tf.record generation

[tanmaypandey7]

Hi, my program is detecting fake labels during tf.record script.
Uploading some screenshot of anaconda:


Uploading images showing error:
a_100.png:

a_101.png:

a_102.png:

a_103.png:

a_104.png:

a_105

a_106

a_107

a_108

a_109

a_110

a_111

a_112

a_113

a_114

a_115

a_116

a_117

a_118

a_119

a_120

a_121

a_122

a_123

a_124

a_125

a_126

I think that it could be due to this weird border surrounding the images. Is it possible to remove this border, if that is causing it?

[vijendra1125]

What is the number of classes you have and what pixel value are you assigning for each class?
Please share your label.pbtxt and create_mask_rcnn_tf_record.py

[tanmaypandey7]

I opened an issue in the PixelAnnotationTool library and they have fixed this issue in the latest version 1.3.2
I will report if i encounter any other problems.

[vijendra1125]

Great! Can you please share link of the issue you opened in the PixelAnnotationTool library.

[tanmaypandey7]

Here you go.
abreheret/PixelAnnotationTool#31

[vijendra1125]

Thank you!