Possibility of missing file. about jsk_apc

[knorth55]

Or you can run make install in instance_occlsegm and build conda environment without ROS.
https://github.com/start-jsk/jsk_apc/blob/master/demos/instance_occlsegm/Makefile

[eagleeye05]

After running make install in instance_occlsegm directory, and as in the sequence of the readme file, I followed by these two commands:

$ cd examples/instance_occlsegm/panoptic_occlusion_segmentation
$ ./train.py --gpu 0 --notrain pix  # instnce-only

And I got this error:

# -----------------------------------------------------------------------------
{'anchor_scales': (4, 8, 16, 32),
 'batch_size': 1,
 'class_names': ['avery_binder',
                 'balloons',
                 'band_aid_tape',
                 'bath_sponge',
                 'black_fashion_gloves',
                 'burts_bees_baby_wipes',
                 'colgate_toothbrush_4pk',
                 'composition_book',
                 'crayons',
                 'duct_tape',
                 'epsom_salts',
                 'expo_eraser',
                 'fiskars_scissors',
                 'flashlight',
                 'glue_sticks',
                 'hand_weight',
                 'hanes_socks',
                 'hinged_ruled_index_cards',
                 'ice_cube_tray',
                 'irish_spring_soap',
                 'laugh_out_loud_jokes',
                 'marbles',
                 'measuring_spoons',
                 'mesh_cup',
                 'mouse_traps',
                 'pie_plates',
                 'plastic_wine_glass',
                 'poland_spring_water',
                 'reynolds_wrap',
                 'robots_dvd',
                 'robots_everywhere',
                 'scotch_sponges',
                 'speed_stick',
                 'table_cloth',
                 'tennis_ball_container',
                 'ticonderoga_pencils',
                 'tissue_box',
                 'toilet_brush',
                 'white_facecloth',
                 'windex'],
 'dataset': 'occlusion',
 'git_hash': '6ed93840',
 'gpu': 0,
 'hostname': 'BQ-DX1100-CT',
 'lr': 0.00125,
 'lr_base': 0.00125,
 'mask_loss': 'softmax',
 'max_epoch': 120,
 'max_size': 1000,
 'min_size': 600,
 'model': 'resnet50',
 'multi_node': False,
 'n_gpu': 1,
 'n_node': 1,
 'noaugmentation': False,
 'notrain': 'pix',
 'out': '/home/student2/PycharmProjects/masterproject/relook/jsk_apc/demos/instance_occlsegm/examples/instance_occlsegm/panoptic_occlusion_segmentation/logs/20201217_035948.655918',
 'pix_loss_scale': 1.0,
 'pooling_func': 'align',
 'pretrained_model': None,
 'rpn_dim': 512,
 'seed': 0,
 'step_size': [80.0, 106.66666666666666],
 'timestamp': '2020-12-17T03:59:48.655918',
 'weight_decay': 0.0001}
# -----------------------------------------------------------------------------
/home/student2/anaconda3/envs/relook/lib/python3.8/site-packages/imgaug/imgaug.py:184: DeprecationWarning: Function `copy_random_state()` is deprecated. Use `imgaug.random.copy_generator_unless_global_rng` instead.
  warn_deprecated(msg, stacklevel=3)
/home/student2/anaconda3/envs/relook/lib/python3.8/site-packages/imgaug/imgaug.py:184: DeprecationWarning: Function `copy_random_state()` is deprecated. Use `imgaug.random.copy_generator_unless_global_rng` instead.
  warn_deprecated(msg, stacklevel=3)
/home/student2/anaconda3/envs/relook/lib/python3.8/site-packages/imgaug/imgaug.py:184: DeprecationWarning: Function `copy_random_state()` is deprecated. Use `imgaug.random.copy_generator_unless_global_rng` instead.
  warn_deprecated(msg, stacklevel=3)
Exception in main training loop: Extension implementation must override __call__.
Traceback (most recent call last):
  File "/home/student2/anaconda3/envs/relook/lib/python3.8/site-packages/chainer/training/trainer.py", line 319, in run
    entry.extension(self)
  File "/home/student2/anaconda3/envs/relook/lib/python3.8/site-packages/chainer/training/extension.py", line 64, in __call__
    raise NotImplementedError(
Will finalize trainer extensions and updater before reraising the exception.
Traceback (most recent call last):
  File "./train.py", line 386, in <module>
    main()
  File "./train.py", line 382, in main
    trainer.run()
  File "/home/student2/anaconda3/envs/relook/lib/python3.8/site-packages/chainer/training/trainer.py", line 349, in run
    six.reraise(*exc_info)
  File "/home/student2/anaconda3/envs/relook/lib/python3.8/site-packages/six.py", line 703, in reraise
    raise value
  File "/home/student2/anaconda3/envs/relook/lib/python3.8/site-packages/chainer/training/trainer.py", line 319, in run
    entry.extension(self)
  File "/home/student2/anaconda3/envs/relook/lib/python3.8/site-packages/chainer/training/extension.py", line 64, in __call__
    raise NotImplementedError(
NotImplementedError: Extension implementation must override __call__.

Any idea how to circumvent this error?

I changed the chainer version to 5.0.0. It works.

[knorth55]

Yes, probably you are right.
It seems demos/instance_occlsegm/examples/instance_occlsegm/semantic_segmentation/view_dataset.py is missing, which is written in https://github.com/start-jsk/jsk_apc/tree/master/demos/instance_occlsegm#without-ros
https://github.com/start-jsk/jsk_apc/tree/master/demos/instance_occlsegm/examples/instance_occlsegm/semantic_segmentation

However, it seems it is just a typo, and we can run it with occlusion_segmentation instead of semantic_segmentation

# Dataset for semantic (image-level) occlusion segmentation
cd examples/instance_occlsegm/instance_segmentation
./view_dataset.py

@wkentaro am I right?

[wkentaro]

Yes, this is a typo ー I sent a fix here: #2728
Thanks for catching and answering this, both!

[IslamFadl]

Please also note that both files "view_dataset.py" , in the README.md ,need another two lines of code:
import sys
sys.path.insert(1, '/home/student2/PycharmProjects/masterproject/relook/jsk_apc/demos/instance_occlsegm')

in order to execute the following line:
from instance_occlsegm_lib.contrib import instance_occlsegm

otherwise a Module not found error appears and it cannot fetch the /instance_occlsegm dir

I suggest it as a temporary solutions, it is not the best style to circumvent the problem, but it works. Otherwise better structure of the code in this file would be needed.

Thank you very much for confirming.

[knorth55]

Please also note that both files "view_dataset.py" , in the README.md ,need another two lines of code:
import sys
sys.path.insert(1, '/home/student2/PycharmProjects/masterproject/relook/jsk_apc/demos/instance_occlsegm')

if you build this package with catkin build, you don't need to set the path because it is ROS package.
but if you want to try it without ROS, you need.

[IslamFadl]

Oh, that's why. I actually didn't install ROS and attempted the 'without ROS' examples thinking it would not affect the code.

Thanks a lot.

[IslamFadl]

After running make install in instance_occlsegm directory, and as in the sequence of the readme file, I followed by these two commands:

$ cd examples/instance_occlsegm/panoptic_occlusion_segmentation
$ ./train.py --gpu 0 --notrain pix  # instnce-only

And I got this error:

# -----------------------------------------------------------------------------
{'anchor_scales': (4, 8, 16, 32),
 'batch_size': 1,
 'class_names': ['avery_binder',
                 'balloons',
                 'band_aid_tape',
                 'bath_sponge',
                 'black_fashion_gloves',
                 'burts_bees_baby_wipes',
                 'colgate_toothbrush_4pk',
                 'composition_book',
                 'crayons',
                 'duct_tape',
                 'epsom_salts',
                 'expo_eraser',
                 'fiskars_scissors',
                 'flashlight',
                 'glue_sticks',
                 'hand_weight',
                 'hanes_socks',
                 'hinged_ruled_index_cards',
                 'ice_cube_tray',
                 'irish_spring_soap',
                 'laugh_out_loud_jokes',
                 'marbles',
                 'measuring_spoons',
                 'mesh_cup',
                 'mouse_traps',
                 'pie_plates',
                 'plastic_wine_glass',
                 'poland_spring_water',
                 'reynolds_wrap',
                 'robots_dvd',
                 'robots_everywhere',
                 'scotch_sponges',
                 'speed_stick',
                 'table_cloth',
                 'tennis_ball_container',
                 'ticonderoga_pencils',
                 'tissue_box',
                 'toilet_brush',
                 'white_facecloth',
                 'windex'],
 'dataset': 'occlusion',
 'git_hash': '6ed93840',
 'gpu': 0,
 'hostname': 'BQ-DX1100-CT',
 'lr': 0.00125,
 'lr_base': 0.00125,
 'mask_loss': 'softmax',
 'max_epoch': 120,
 'max_size': 1000,
 'min_size': 600,
 'model': 'resnet50',
 'multi_node': False,
 'n_gpu': 1,
 'n_node': 1,
 'noaugmentation': False,
 'notrain': 'pix',
 'out': '/home/student2/PycharmProjects/masterproject/relook/jsk_apc/demos/instance_occlsegm/examples/instance_occlsegm/panoptic_occlusion_segmentation/logs/20201217_035948.655918',
 'pix_loss_scale': 1.0,
 'pooling_func': 'align',
 'pretrained_model': None,
 'rpn_dim': 512,
 'seed': 0,
 'step_size': [80.0, 106.66666666666666],
 'timestamp': '2020-12-17T03:59:48.655918',
 'weight_decay': 0.0001}
# -----------------------------------------------------------------------------
/home/student2/anaconda3/envs/relook/lib/python3.8/site-packages/imgaug/imgaug.py:184: DeprecationWarning: Function `copy_random_state()` is deprecated. Use `imgaug.random.copy_generator_unless_global_rng` instead.
  warn_deprecated(msg, stacklevel=3)
/home/student2/anaconda3/envs/relook/lib/python3.8/site-packages/imgaug/imgaug.py:184: DeprecationWarning: Function `copy_random_state()` is deprecated. Use `imgaug.random.copy_generator_unless_global_rng` instead.
  warn_deprecated(msg, stacklevel=3)
/home/student2/anaconda3/envs/relook/lib/python3.8/site-packages/imgaug/imgaug.py:184: DeprecationWarning: Function `copy_random_state()` is deprecated. Use `imgaug.random.copy_generator_unless_global_rng` instead.
  warn_deprecated(msg, stacklevel=3)
Exception in main training loop: Extension implementation must override __call__.
Traceback (most recent call last):
  File "/home/student2/anaconda3/envs/relook/lib/python3.8/site-packages/chainer/training/trainer.py", line 319, in run
    entry.extension(self)
  File "/home/student2/anaconda3/envs/relook/lib/python3.8/site-packages/chainer/training/extension.py", line 64, in __call__
    raise NotImplementedError(
Will finalize trainer extensions and updater before reraising the exception.
Traceback (most recent call last):
  File "./train.py", line 386, in <module>
    main()
  File "./train.py", line 382, in main
    trainer.run()
  File "/home/student2/anaconda3/envs/relook/lib/python3.8/site-packages/chainer/training/trainer.py", line 349, in run
    six.reraise(*exc_info)
  File "/home/student2/anaconda3/envs/relook/lib/python3.8/site-packages/six.py", line 703, in reraise
    raise value
  File "/home/student2/anaconda3/envs/relook/lib/python3.8/site-packages/chainer/training/trainer.py", line 319, in run
    entry.extension(self)
  File "/home/student2/anaconda3/envs/relook/lib/python3.8/site-packages/chainer/training/extension.py", line 64, in __call__
    raise NotImplementedError(
NotImplementedError: Extension implementation must override __call__.

Any idea how to circumvent this error?

[wkentaro]

We developed this with older Chainer, so maybe you have to check its version or modify the Extension to adapt to the latest version.

[IslamFadl]

Thanks for the response.
I am a bit confused, which CuPy pakage do you have installed in this specific case?

[eagleeye05]

Thanks for the response.
I am a bit confused, which CuPy pakage do you have installed in this specific case?

I download the cupy-cuda101. It works.