aasimsani / artificial_manga_panel_dataset Goto Github PK

In the Kaggle datasets/speech_bubbles_dataset/writing_area_labels.csv file, the 'imagename' entries contain a "~" character in their paths, whereas the actual file paths in the speech_bubbles_dataset/files/ directory use a "-" character instead.

wrong: "datasets/speech_bubbles_dataset/files/white~friendlystock-chat-bubble-47.png"
correct: "datasets/speech_bubbles_dataset/files/white-friendlystock-chat-bubble-47.png"

As a workaround, I've implemented a simple string replacement operation in my local script to replace the "~" character with "-", and the function is working as expected.
main.py 128 and 191 you can add, after the initial speech_bubble_tags declaration so that:

speech_bubble_tags = pd.read_csv(speech_bubbles_path +
                                         "writing_area_labels.csv")
# Replace "~" with "-" in the 'imagename' column to match the file paths
speech_bubble_tags['imagename'] = speech_bubble_tags['imagename'].str.replace('~', '-') # fix kaggle path error

Error logs:
Testing:
FAILURES test_speech_bubble_dumping
speech_bubble_writing_area = speech_bubble_tags[area_idx]['label']

      speech_bubble_writing_area = speech_bubble_writing_area.values[0]

E IndexError: index 0 is out of bounds for axis 0 with size 0

preprocesing/layout_engine/page_dataset_creator.py:1546: IndexError

Traceback (most recent call last):
File "main.py", line 203, in
speech_bubble_tags
File "/home/john/Desktop/translate/artificial_manga_panel_dataset/preprocesing/layout_engine/page_dataset_creator.py", line 2365, in create_page_metadata
speech_bubble_tags
File "/home/john/Desktop/translate/artificial_manga_panel_dataset/preprocesing/layout_engine/page_dataset_creator.py", line 1660, in populate_panels
minimum_speech_bubbles
File "/home/john/Desktop/translate/artificial_manga_panel_dataset/preprocesing/layout_engine/page_dataset_creator.py", line 1546, in create_single_panel_metadata
speech_bubble_writing_area = speech_bubble_writing_area.values[0]
IndexError: index 0 is out of bounds for axis 0 with size 0

P.S. This project is really awesome thank you!

Hello! I'm Felipe.

Do you have a pretrained IA for Object Detection for Comics using Manga109 Annotations?
I read README.md and you used this paper to benchmark your dataset.

I want to create a autogenerated videos from Manga or Manhwa (get panels, show panels instead whole page, get text of bubbles speech, create narrations, add sfx, etc.). The idea, you will input manga's page and it will return a video

Disclaimer: I don't too much about IA. I'm a frontend developer and its my first time at IA

Conclusion: If you don't have a pretrained or source code for implementation, would you like to give me an idea to start build my own?