If no mask file, then file browser text also be removed.
Ultimately, put an image / mask preview in file browser(in future version)

It's somewhat confusing.

• How about open manga project directory in explorer(Windows) right after create new project directory?

or...?

• Rename id, objectNames along to consistent convention
  • Use camel case
  • Specify type in tail (Image -> XxxImg)
• Decompose szmc-0.1.0.qml into multiple modules

--
state

• window.state

dialog

• MessageDialog - id: msgDialog
• FileDialog - id: projectOpenDialog
• Action - id: openProject

toolButton

....
more

references:
https://doc.qt.io/qt-5/qtqml-javascript-expressions.html
https://doc.qt.io/qt-5/qtqml-javascript-hostenvironment.html#

and think about functional reactive programming with qml...

If user press {gen_mask, rmtxt, restore_prev_image, ...} after the manga project folder is changed from outside (delete folder, files, etc..), exception occurs:

FileNotFoundError: [Errno 2] No such file or directory: '...'

• rmtxt on not existing image (damaged images or masks)
• gen_mask on not existing image (damaged images)
• restore_prev_image but no prev_image (damaged prev_images)

Find all of this kind of crashes, and warn to user that manga project is damaged.
Or automatically fix.. do something..

Professional editors commonly use digitizer(graphics tablet).
SZMC gui need to support it.

Write Test Cases First!
https://jaku.tistory.com/17

• Linux
• Windows
  --
• no extension images OK

Automate if you can..

----------- coverage: platform linux, python 3.7.3-final-0 -----------
Name Stmts Miss Cover Missing

/home/kur/dev/SickZil-Machine/src/ImListModel.py 31 3 90% 35-36, 47
/home/kur/dev/SickZil-Machine/src/init.py 0 0 100%
/home/kur/dev/SickZil-Machine/src/config.py 27 1 96% 52
/home/kur/dev/SickZil-Machine/src/core.py 91 19 79% 55-58, 62-69, 141-143, 147-154
/home/kur/dev/SickZil-Machine/src/gui.py 98 34 65% 19-20, 79-91, 96-99, 103-106, 110-112, 142-154, 158-170
/home/kur/dev/SickZil-Machine/src/imgio.py 25 0 100%
/home/kur/dev/SickZil-Machine/src/main.py 9 9 0% 8-19
/home/kur/dev/SickZil-Machine/src/state.py 53 13 75% 41-51, 84-85, 89, 92
/home/kur/dev/SickZil-Machine/src/utils/init.py 0 0 100%
/home/kur/dev/SickZil-Machine/src/utils/fp.py 58 27 53% 16, 19, 26, 30-31, 34-37, 49, 66-89
/home/kur/dev/SickZil-Machine/src/utils/futils.py 46 20 57% 19-24, 33-35, 58-68
/home/kur/dev/SickZil-Machine/src/utils/imutils.py 112 35 69% 18, 33-34, 43-44, 62-64, 69-71, 81-85, 155, 158, 177-186, 199-210, 216-217

TOTAL 550 161 71%

64bit, gpu/cpu supported.

• Press backspace or delete keys to remove current mask
• select images in image list and deletion all selected images

• Async gen_mask
• Async rmtxt
  --
• Async gen_mask_all
• Async rmtxt_all
• User can edit mask while xx_all
• display which image & mask are processed

Also image list is too long, some weird interaction occurs in list view(Maybe another issue).

Write log file in release build.
How user to report to dev? Think about it..

• Create a log file each time you run SZMC
• Log file name is time(or something sortable)
• Print all informations
  --
• Maintain server Use cloud service(new relic, firebase, sentry, etc..)
• Notify users to collect data
• Send log to server
  --
• Test error cases: does client send error log correctly? when
  • Crash
  • Invalid machine
    etc..

https://stackoverflow.com/questions/44853059/tensorflow-logging-messages-do-not-appear

Create web page for v0.1.1

• Web site for SZMC: Release / Development / Research

For Release and Information

• Blog(Move posts from old private blog)
• Download SZMC

For Contribution

• Contribute Dataset
• Donate

For Research

• Download Dataset
• Challenges
  • SegNet Challenge
  • ComplNet Challenge
  • SZMC Challenge

Others

• link to Github repositories
• Contact

User can create from "flat image directory"(just has some image files) to "manga project" directory for SickZil-Machine
"manga project" directory has {images, masks, origin_imgs, origin_segmaps}.
UI

• This function is activated when user selects "flat image directory" in projectOpenDialog.
• Inform the user what will happen before creation using warning dialog.
• User can edit name of "manga project directory" using interactive dialog.
• Decide default name of manga project directory.

internals

• Test inter-device copying images to manga project directory
• If project directory already exists, then copy not existing images(no overwirte)

For better data gathering, I suggest user data collection

It will be okay If it can be chosen by user with configuration.

• Save original images to other directory
  If user change image right affter revert, warn user to overwrite prev image?
• Save previous mask to prev_masks
  --
• Revert to original image
• Revert to prev mask / or first mask? and mask: union with current mask?
• If no mask file, then just reset mask(to empty)

Just save/revert only images in version 0.1.0

Reversion is very limited undo/redo functionality.
In later versions, implement a proper undo / redo system.

It's too tedious.

How to test released executable??
How to test qml qui?

Set rectangle area with mouse, and then seg->compl.
it would be very efficient. And very useful when user edit a mask after full automated masks generation.

• User can select part of mask to inference

Brush

• cursor
• size control
• tool button
• shortcut key
  slider for size control later.

Rectangle

• cursor
• drawing effect (like gimp)
• tool button
• actual drawing
• shortcut key

Line

Line(use shift, like gimp)
draw line with shift key (like gimp)
drawing effect (like gimp)

later.

Panning

• cursor
• tool button
• Panning image with mouse wheel(Is it possible?)

Use right mouse to erase? Think about it..

• 디시인사이드 만화 갤러리
• Add Request(부탁): please inform szmc to use icon.
• Add escape clause(면책 조항)
• Rework release page and release section in README
  --
• Correct English docs
• Reddit
  --
  etc..

🤔🤔🤔

That would be better...
Also think about panning button inconsistency(buttons are have to be clickable)

Do it after decompsing qml to multiple modules(refactoring)

Like #35, user can revert part of selected image. It would be very convenient too.

Currently, the only way to solve this problem is to turn the program off and on.

1. requirements.txt
   Add a requirements.txt for dependencies such as Tensorflow with an exact required version

2. black linting/formatting
   Add a Black formatting configuration to ensure code quality https://github.com/psf/black

3. License
   Add a LICENSE file

4. Don't put actual initiation code under under if __name__ == '__main__':, instead you should:

def main():
    consts.load_config('../resource/config.json')
    app = QApplication(sys.argv)

    engine = QQmlApplicationEngine()
    main_window = gui.MainWindow(engine)
    
    sys.exit(app.exec_())

if __name__ == '__main__':
    main()

Problem

If image is too big, SZMC crashes.

• Catch OOM,
• and then update seg_limit or compl_limit in config.json

Generate Masks:  45%|████████████████████████████▌                                   | 127/285 [10:15<12:25,  4.72s/it]Traceback (most recent call last):
  File "site-packages\tensorflow\python\client\session.py", line 1356, in _do_call
  File "site-packages\tensorflow\python\client\session.py", line 1341, in _run_fn
  File "site-packages\tensorflow\python\client\session.py", line 1429, in _call_tf_sessionrun
tensorflow.python.framework.errors_impl.ResourceExhaustedError: 2 root error(s) found.
  (0) Resource exhausted: OOM when allocating tensor with shape[1,32,2912,1280] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc
         [[{{node snet/concatenate_4/concat}}]]
Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.

         [[snet/conv2d_19/truediv/_5]]
Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.

  (1) Resource exhausted: OOM when allocating tensor with shape[1,32,2912,1280] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc
         [[{{node snet/concatenate_4/concat}}]]
Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.

0 successful operations.
0 derived errors ignored.

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "src\gui.py", line 178, in gen_mask_all
  File "site-packages\tqdm\_tqdm.py", line 1017, in __iter__
  File "src\gui.py", line 20, in imgpath2mask
  File "src\utils\fp.py", line 5, in <lambda>
  File "site-packages\funcy\funcs.py", line 106, in <lambda>
  File "site-packages\funcy\funcs.py", line 106, in <lambda>
  File "src\core.py", line 104, in segmap
  File "src\utils\fp.py", line 5, in <lambda>
  File "site-packages\funcy\funcs.py", line 106, in <lambda>
  File "site-packages\funcy\funcs.py", line 106, in <lambda>
  File "site-packages\funcy\funcs.py", line 106, in <lambda>
  [Previous line repeated 1 more time]
  File "src\core.py", line 103, in <lambda>
  File "src\core.py", line 59, in segment
  File "src\core.py", line 37, in segment_or_oom
  File "src\core.py", line 97, in snet
  File "site-packages\tensorflow\python\client\session.py", line 950, in run
  File "site-packages\tensorflow\python\client\session.py", line 1173, in _run
  File "site-packages\tensorflow\python\client\session.py", line 1350, in _do_run
  File "site-packages\tensorflow\python\client\session.py", line 1370, in _do_call
tensorflow.python.framework.errors_impl.ResourceExhaustedError: 2 root error(s) found.
  (0) Resource exhausted: OOM when allocating tensor with shape[1,32,2912,1280] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc
         [[node snet/concatenate_4/concat (defined at src\core.py:25) ]]
Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.

         [[snet/conv2d_19/truediv/_5]]
Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.

  (1) Resource exhausted: OOM when allocating tensor with shape[1,32,2912,1280] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc
         [[node snet/concatenate_4/concat (defined at src\core.py:25) ]]
Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.

0 successful operations.
0 derived errors ignored.

Original stack trace for 'snet/concatenate_4/concat':
  File "src\main.py", line 11, in <module>
  File "<frozen importlib._bootstrap>", line 971, in _find_and_load
  File "<frozen importlib._bootstrap>", line 955, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 665, in _load_unlocked
  File "c:\users\kur\appdata\local\programs\python\python36\lib\site-packages\PyInstaller\loader\pyimod03_importers.py", line 627, in exec_module
  File "src\gui.py", line 12, in <module>
  File "<frozen importlib._bootstrap>", line 971, in _find_and_load
  File "<frozen importlib._bootstrap>", line 955, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 665, in _load_unlocked
  File "c:\users\kur\appdata\local\programs\python\python36\lib\site-packages\PyInstaller\loader\pyimod03_importers.py", line 627, in exec_module
  File "src\core.py", line 27, in <module>
  File "src\core.py", line 25, in load_model
  File "site-packages\tensorflow\python\util\deprecation.py", line 507, in new_func
  File "site-packages\tensorflow\python\framework\importer.py", line 443, in import_graph_def
  File "site-packages\tensorflow\python\framework\importer.py", line 236, in _ProcessNewOps
  File "site-packages\tensorflow\python\framework\ops.py", line 3751, in _add_new_tf_operations
  File "site-packages\tensorflow\python\framework\ops.py", line 3751, in <listcomp>
  File "site-packages\tensorflow\python\framework\ops.py", line 3641, in _create_op_from_tf_operation
  File "site-packages\tensorflow\python\framework\ops.py", line 2005, in __init__

If the machine crashed while generating the masks / removing texts by GPU OOM error, log in cmd would be like above.

Solution

before

after

Please edit seg_limit in resource/config.json to smaller value if crash occurs in generating masks.
Please edit compl_limit in resource/config.json to smaller value if crash occurs in removing texts.

In most system, "seg_limit" = 3000000 will fix the bug.
If not, decrease seg_limit more.

Other Solution

Wait next release... It will be fixed, and automatically set the limits.

mproj -> mproj(1) -> mproj(2) ...

Good for hackers like you and me!

• Convert many "flat img dir" -> "mproj"
• Open many mproj
• full-auto rmTxt on many mproj

... and more..

README.md says
Unzip the release file and copy SickZil-Machine-0.1.1-pre0-win64-cpu-eng/resource/cnet and SickZil-Machine-0.1.1-pre0-win64-cpu-eng/resource/snet directories to SickZil-Machine/resource.
But I can't find those files in the release zip. At least not in SickZil-Machine-0.1.1-pre0-win64-cpu-eng.zip or SickZil-Machine-0.1.1-pre0-win64-gpu-eng.zip.

I've found a bug that the application loads only one page of image when I load a project after loading a project that contains of image.

another case, entire directory cannot be loaded, It seems several glitches exists on file checking routine for filtering images or non-images.

• Button tool tips: explain function, shortcut, etc..
• Edit typo
• set Pen / Eraser mode cursor colors or something?
• Status bar later
• multilingual support preparation
  • check: pyinstaller can set default cmd arg? or..?
• Y key to change to previous tool

--

• Read config.json file and use it (both in qml, py)
• Full Screen
• Enter key for single rmtxt
• Display current working project(display with title)
• gen_mask, rmtxt progress bar later

--

• Fix bug: press up/down key in startup page, it deletes qml file in some system. (#16 )
• next/prev button useless

• Warn to User: It takes very long time! (expected time: ...) Ok / No
• Change icon order: XX-all first, vertical bar, simgle image tools...
• mask_all
• rmtxt_all
  --
  if there is mask already then rmtxt?

Me is an Apple machine, macbook pro user needs it!
hope can see it in the near future!

이미지 첨부해봅니다
마스크를 그리는 작업이던 지우는 작업이던
진행도중(좌클릭 상태로) 파일을 넘기면 바로 그려지는 작업으로 전환
다른 방향에서 진행중인 경우 다음장 첫화면 보이는곳까지 직선 쫙

그냥... 이런게 있길래요...
근데 여기다가 자꾸 불만사항 적어도 되는지요; 죄송스럽네요

Someone thinks it sucks.

If image/mask is unchanged(inside and outside too), then do not read from disk. Instead use cache in memory.

apply only mask editing, not loading mask

이미 있는거였네

_{(from https://gall.dcinside.com/board/view/?id=comic_new1&no=10709824)}

If SZMC currently display above image, user press RmTxtAll or RmTxt
then SZMC can't load text removed image after rm_txt.

what..? why?

RmTxtAll, RmTxt both have this bug
This bug is fault of gui, not the core.

Tested on manjaro 18, i7, gtx1070ti.

For saving logs, image files, etc.. from client

Add sending logs/imgs features to client too!(open another issues)

Client sends

• images, masks, prev_images
• User Statistics
  • How many people uses SZMC?
  • System Information
  • Running time for h x w image
  • Use Pattern
• Bug, Crash report

• imgio.save
• unicode path
  etc..

Find optimal size of image input of snet/cnet.
It can be varied by OS, GPU/CPU, RAM size, etc...
Plot relation of size and processing time.
If there is no general relation, try to implement "Dynamic" snet/cnet processing time optimization

• Save optimized h/w of snet/cnet processing (save config file)
• Save processing time
• Calculate expected time using saved processing time, h, w

The first image takes longer, the second faster.

• Is it possible even if the first image is small? Can this improve performance?
• Monitor gpu usage in windows / linux

Issue Templates

• Bug Report (Kor)
• Bug Report (Eng)
• Feature Request (Kor)
• Feature Request (Eng)

contact

• Open public email for SZMC

수십장 넣고 대패질후 검수를 할때 일일이 마우스 클릭 중입니다
페이지 업다운으로 사진 전환이 가능했으면 좋겠습니다
위아래 방향키로 이미지가 안넘어가요..
계속 좌상단 열기만 나옵니다
ver. cpu

이미지 크기가 크면 한 화면에 잡을 수 있는 내용이 너무 적어서 마스크 그리기/지우기에 어려움이 있네요.
확대/축소 기능이 있으면 좀 더 이용하기 편할 것 같습니다.

좋은 프로그램 감사합니다

Feature Request (Kor)

만화갤러리에서 글을 보고 제안하기 위해 글을 남겨봅니다.
몇 가지 찾아본 결과 무료 일본어 ocr로 nhocr이 있고,
파파고에서도 파이썬을 위한 무료 API를 제공하는 것으로 보입니다. 다만 하루 1만자의 글자수 제한이 있고 유료 이용시 100만자당 월 2만원이 든다고 합니다.링크
유료 API를 제안드리긴 힘들 것 같고 ocr로 글씨의 좌표를 찾아 일본어를 추출하고 삽입할 번역내용을 입력하는 부분은 가능하다고 생각됩니다. 추출한 내용을 띄우는 일본어 텍스트 박스와 번역내용을 입력할 텍스트 박스를 띄우면 되지 않을까 싶습니다.
다만 글씨크기 조절에 대해서는 아이디어가 없네요. 이 부분은 죄송합니다.
바쁘시겠지만 다음 버전에서 고려해주시면 감사하겠습니다.

• Tutorial in startup image Useless
• Korean version -> (#23)
• Create homepage(githun.io) for project -> (#24)
• Complete README.md(gif, how it works, 어케했 노-> 누)

--

• Button and Key manual in startup page - Especially, keyboard buttons.
• Introduction Video on Youtube <- Just Do t!
• limits(pool color..) and requirements.txt

--

• Kor
• Eng

And select lang.json in config.json.
Study about i18n.

This bug only occurs in my desktop(Manjaro 18.0.4)
I don't know why it occurs.... maybe bug of qt.