COCO Panoptic data does not have 'color' attribute about panopticapi

[alexander-kirillov]

You should use panoptic_coco_categories.json from this repo for --categories_json_file. The json file has the right color field.

[Robotatron]

@alexander-kirillov

You should use panoptic_coco_categories.json from this repo for --categories_json_file. The json file has the right color field.

Do we have to modify the panoptic_coco_categories.json if we are using a custom dataset with different labels? I am getting an error Exception: Panoptic coco categories file does not contain category with id: 26

[alexander-kirillov]

Thanks you for spotting the bug. The code is fixed now. We'll also update annotations later today to add colors there as well.

[cao-nv]

Have you fixed this issue?
I got this problem today. Both panopticapi and annotations files are the latest releases

[alexander-kirillov]

There is a delay with the update due to a migration of the COCO storage. I'm sorry! Nevertheless, current data is correct. New files just have the color field and new segment ID encoding, so there is no need in waiting for them now.

You can use panoptic_coco_categories.json file that has the right color for each category. Also, format_converter.py has been fixed and does not require the color field anymore.

[cao-nv]

Sorry if I was wrong, but I found format_converter.py still used keywords color here.

panopticapi/format_converter.py

Line 26 in e30ca01

self.taken_colors.add(tuple(category['color']))

This is my command to convert the format, the categories.json was extracted from the annotation json file of the training data.

(.global_env) cao@slspGPU1:~/workspace/data/COCO$ python panopticapi/format_converter.py --source_folder='../../COCO2018/panoptic/val' --images_json_file='annotations/panoptic_val2017.json' --categories_json_file='./annotations/categories.json' --segmentations_folder='val_seg' --predictions_json_file='val.json'

Output of the command! Thanks

Reading image set information from annotations/panoptic_val2017.json
CONVERTING...
2 channels PNG panoptic format:
        Source folder: ../../COCO2018/panoptic/val
TO
COCO format:
        Segmentation folder: val_seg
        JSON file: val.json
Number of cores: 8, images per core: 625
Core: 0, 0 from 625 images converted
Core: 1, 0 from 625 images converted
Core: 2, 0 from 625 images converted
Core: 3, 0 from 625 images converted
Core: 4, 0 from 625 images converted
Core: 5, 0 from 625 images converted
Core: 6, 0 from 625 images converted
Caught exception in worker thread:
Traceback (most recent call last):
  File "/mnt/hdd1/cao/workspace/data/COCO/panopticapi/utils.py", line 12, in wrapper
    return f(*args, **kwargs)
  File "panopticapi/format_converter.py", line 69, in convert_single_core
    color_gererator = ColorGenerator(categories)
  File "panopticapi/format_converter.py", line 26, in __init__
    self.taken_colors.add(tuple(category['color']))
KeyError: u'color'
Traceback (most recent call last):
  File "panopticapi/format_converter.py", line 156, in <module>
    args.void)
  File "panopticapi/format_converter.py", line 125, in converter
    annotations.extend(p.get())
  File "/usr/lib/python2.7/multiprocessing/pool.py", line 567, in get
    raise self._value
KeyError: u'color'
Core: 7, 0 from 625 images converted
Caught exception in worker thread:
Traceback (most recent call last):
  File "/mnt/hdd1/cao/workspace/data/COCO/panopticapi/utils.py", line 12, in wrapper
    return f(*args, **kwargs)
  File "panopticapi/format_converter.py", line 69, in convert_single_core
    color_gererator = ColorGenerator(categories)
  File "panopticapi/format_converter.py", line 26, in __init__
    self.taken_colors.add(tuple(category['color']))
KeyError: u'color'
Caught exception in worker thread:
Traceback (most recent call last):
  File "/mnt/hdd1/cao/workspace/data/COCO/panopticapi/utils.py", line 12, in wrapper
    return f(*args, **kwargs)
  File "panopticapi/format_converter.py", line 69, in convert_single_core
    color_gererator = ColorGenerator(categories)
  File "panopticapi/format_converter.py", line 26, in __init__
    self.taken_colors.add(tuple(category['color']))
KeyError: u'color'
Caught exception in worker thread:
Traceback (most recent call last):
  File "/mnt/hdd1/cao/workspace/data/COCO/panopticapi/utils.py", line 12, in wrapper
    return f(*args, **kwargs)
  File "panopticapi/format_converter.py", line 69, in convert_single_core
    color_gererator = ColorGenerator(categories)
  File "panopticapi/format_converter.py", line 26, in __init__
    self.taken_colors.add(tuple(category['color']))
KeyError: u'color'
Caught exception in worker thread:
Traceback (most recent call last):
  File "/mnt/hdd1/cao/workspace/data/COCO/panopticapi/utils.py", line 12, in wrapper
    return f(*args, **kwargs)
  File "panopticapi/format_converter.py", line 69, in convert_single_core
    color_gererator = ColorGenerator(categories)
  File "panopticapi/format_converter.py", line 26, in __init__
    self.taken_colors.add(tuple(category['color']))
KeyError: u'color'
Caught exception in worker thread:
Traceback (most recent call last):
  File "/mnt/hdd1/cao/workspace/data/COCO/panopticapi/utils.py", line 12, in wrapper
    return f(*args, **kwargs)
  File "panopticapi/format_converter.py", line 69, in convert_single_core
    color_gererator = ColorGenerator(categories)
  File "panopticapi/format_converter.py", line 26, in __init__
    self.taken_colors.add(tuple(category['color']))
KeyError: u'color'
Caught exception in worker thread:
Traceback (most recent call last):
  File "/mnt/hdd1/cao/workspace/data/COCO/panopticapi/utils.py", line 12, in wrapper
    return f(*args, **kwargs)
  File "panopticapi/format_converter.py", line 69, in convert_single_core
    color_gererator = ColorGenerator(categories)
  File "panopticapi/format_converter.py", line 26, in __init__
    self.taken_colors.add(tuple(category['color']))
KeyError: u'color'
Caught exception in worker thread:
Traceback (most recent call last):
  File "/mnt/hdd1/cao/workspace/data/COCO/panopticapi/utils.py", line 12, in wrapper
    return f(*args, **kwargs)
  File "panopticapi/format_converter.py", line 69, in convert_single_core
    color_gererator = ColorGenerator(categories)
  File "panopticapi/format_converter.py", line 26, in __init__
    self.taken_colors.add(tuple(category['color']))
KeyError: u'color'

[cao-nv]

sorry for my mistake, and thanks for you support!

[eliethesaiyan]

@alexander-kirillov , the panoptic_annotation(panoptic_train2017 and val) still have no color attribute when using panoptic2detection_coco_format.py , does the detect task needs color in the annotations? by looking in sample data , panoptic_coco_detection_format.json seems to have the color attribute, would you please guide me to the right Panoptic_annotation files(with color attribute), Thank you