Table 1
Research phases of this paper.
| NUMBER | RESEARCH PHASE |
|---|---|
| 1 | Training data collection of Four-Panel Cartoons (FPCs). |
| 2 | Labelling process of the training dataset. |
| 3 | Fine-tuning of the YOLOv5 Model. |
| 4 | YOLOv5_FPC model evaluation: F1-score. |
| 5 | Image file collection from the Chosun Ilbo News Library (1920–1940), totaling 47,777 JPG files. |
| 6 | Data mining: Deployment of the YOLOv5_FPC model to the 47,777 JPG files from the Chosun Ilbo News Library, to detect FPC image ojects. |
| 7 | Database curation: Uploading the Excel and CSV file which contains metadata for URLs of YOLOv5_FPC-detected 1035 images files (1040 FPC objects), which includes preciously undiscovered FPCs, to the JOHD Dataverse (Lee etal., 2024a). |
| 8 | Data analysis of the detected FPC objects. |
| 9 | Development of the YOLOv5_FPC-Detector script, leveraging the Google Colab platform for enhanced computational efficiency and wider application for the public. |
Figure 1
The initial YOLOv5 Model could not detect an FPC.
Table 2
Matrix and era of the “Four-panel Cartoon Image Dataset”.
| SET | FPC MATRIX | COLONIAL ERA | POST-COLONIAL ERA |
|---|---|---|---|
| Training | 4 × 1 | 31 | 37 |
| 2 × 2 | 26 | 28 | |
| Validation | 4 × 1 | 12 | 3 |
| 2 × 2 | 4 | 7 | |
| Testing | 4 × 1 | 8 | 6 |
| 2 × 2 | 6 | 7 |
Figure 2
Model performance while fine-tuning.
Figure 3
F1-score of our YOLOv5_FPC model.
Figure 4
Chosun Ilbo News Library newspaper metadata (1920–1940) (ChosunIlboNewsLibrary, 2024).
Figure 5
47,777 image files collected from the Chosun Ilbo News Library (1920–1940) (ChosunIlboNewsLibrary, 2024).
Figure 6
Our Dataset: “Metadata for the YOLOv5_FPC Detected Images” (Lee et al., 2024a) containing the URLs (YOLOv5_FPC-detected 1035 image files; 1040 FPC objects in total), and their publication dates sourced from the Chosun Ilbo News Library (1920–1940).
Figure 7
Previously undiscovered FPC image data from the Chosun Ilbo News Library digital archive (ChosunIlbo, 2024).
Table 3
Metadata definitions of Figure 4 Excel file columns.
| COLUMN NAMES | DEFINITION |
|---|---|
| id | The unique identifier for each article of Chosun Ilbo. |
| page_no | The page number of the article. |
| title | The title of the newspaper article. |
| regdate | The registration date of the article. |
| type | The type of the article. |
| publication_day | The day of the week when the article was published. |
| section | The section of the newspaper where the article is placed. |
| publication_date | The date when the article was published (Year-Month-Day). |
| completeness | The completeness of the article (“Y” indicates Yes). |
| body | The main text of the article. |
| publication_no | The publication number. |
| node_id | Numeric identifier associated with the article. |
| source_image_file | The name of the image file. |
| @timestamp | The timestamp indicating when the newspaper textual data was collected to the Excel file (Figure 4). |
| isn | The International Standard Number (ISN) associated with the article. |
| source_xml_file | The XML file name. |
| page_section | The section of the newspaper (society, general section, advertisement, politics, culture). |
| url | The URL to the article. |
| image | The URLs containing the website links to the scanned and digitized image files of the Chosun Ilbo newspaper database (used for collecting 47,777 JPG image files in this research). |
| sub_title | The subtitle of the article. |
| authors | The author(s) who wrote the newspaper article. |
Table 4
Frequency analysis of the 1040 YOLOv5_FPC-detected FPCs discovered from the Chosun Ilbo News Library (1920–1940) (ChosunIlboNewsLibrary, 2024).
| INDEX | NAME OF FPC DETECTED USING THE YOLOV5_FPC MODEL (CHOSUN ILBO NEWS LIBRARY, SPANNING 1920–1940) | FREQUENCY (PER FPC) |
|---|---|---|
| 1 | Meongteongguri | 726 |
| 2 | Byeokchangho | 126 |
| 3 | Japanese Language-Written Cartoon | 104 |
| 4 | Baekgongsan | 13 |
| 5 | Dolbo and Mikki | 12 |
| 6 | Ttukdugiui Seollori | 11 |
| 7 | Chador’s Adventure | 9 |
| 8 | Arctic Exploration | 8 |
| 9 | Rubber Balloon | 7 |
| 10 | Football Player | 8 |
| 11 | Makdongi and Goose | 2 |
| 12 | Biography of a Fool | 2 |
| 13 | Paengkenggun’s Monkey Catching | 1 |
| 14 | Buffalo and Fish | 1 |
| 15 | Then Yes | 1 |
| 16 | Hobang Bridge | 1 |
| 17 | Ice Snack | 1 |
| 18 | The Evil of Alcohol | 1 |
| 19 | Put Your Hands Up | 1 |
| 20 | Better Radio | 1 |
| 21 | Tiger Den | 1 |
| 22 | Hide and Seek | 1 |
| 23 | Samyeong’s Caramel Cartoon | 1 |
| 24 | Love Trees | 1 |
Table 5
Frequency comparison of the “Meongteongguri” series.
| SERIES OF “MEONGTEONGGURI” FPC | CHUNG’S RESEARCH FINDINGS OF FPCS (CHUNG, 2016) | OUR YOLOV5_FPC FINDINGS OF FPCS | DIFFERENCES (PER FPC) |
|---|---|---|---|
| Reporter Life Part 1 | None | 35 | +35 |
| Modern Life | None | 4 | +4 |
| Social Work | 50 | 62 | +12 |
| Heonmulkyeoji | 48 | 55 | +7 |
| Ssutdeokdaegi | 18 | 19 | +1 |
| Student Life | 12 | 12 | 0 |
| Ssonawatso | 9 | 9 | 0 |
| Self-sufficiency | 87 | 86 | –1 |
| Round the World | 148 | 147 | –1 |
| Hunger Life | 50 | 19 | –31 |
| Dating Life | 181 | 178 | –3 |
| Family Life | 102 | 100 | –2 |
Figure 8
Automatic FPC detection using the weights of the YOLOv5_FPC model on Google Colab. This script imports the weights and downloads dependencies required for the automatic detection process.
Figure 9
Users can simply upload their files to detect FPCs on their local computers.
Figure 10
The detected FPCs are saved on their local computers.