OregairuChar: A Benchmark Dataset for Character Appearance Frequency Analysis in My Teen Romantic Comedy SNAFU

TL;DR

Understanding narrative dynamics in anime through character appearance frequency, especially in stylized domains, is the core problem. The paper introduces OregairuChar, a dataset of 1600 frames with 2860 bounding boxes for $K=11$ main characters from My Teen Romantic Comedy SNAFU Season 3, and provides a baseline demo using YOLOv5 with an embedding-based identity refinement to enable temporal appearance analysis. Key contributions include the high-quality, temporally coherent dataset, comprehensive model evaluation on stylized content, and a demonstration of automated appearance-frequency analysis across episodes. This resource enables computational narrative studies, supporting development of temporally aware, character-centric detection and deeper exploration of narrative dynamics in stylized media.

Abstract

The analysis of character appearance frequency is essential for understanding narrative structure, character prominence, and story progression in anime. In this work, we introduce OregairuChar, a benchmark dataset designed for appearance frequency analysis in the anime series My Teen Romantic Comedy SNAFU. The dataset comprises 1600 manually selected frames from the third season, annotated with 2860 bounding boxes across 11 main characters. OregairuChar captures diverse visual challenges, including occlusion, pose variation, and inter-character similarity, providing a realistic basis for appearance-based studies. To enable quantitative research, we benchmark several object detection models on the dataset and leverage their predictions for fine-grained, episode-level analysis of character presence over time. This approach reveals patterns of character prominence and their evolution within the narrative. By emphasizing appearance frequency, OregairuChar serves as a valuable resource for exploring computational narrative dynamics and character-centric storytelling in stylized media.

Figures (5)

• Figure 1: Example annotation in OregairuChar. A sample frame from My Teen Romantic Comedy SNAFU with bounding boxes for multiple main characters.
• Figure 2: Bounding‐box counts per character in OregairuChar. The plot highlights the protagonist Hachiman Hikigaya’s dominance and the long-tail presence of supporting roles.
• Figure 3: Challenges in character appearance frequency analysis illustrated by OregairuChar.
• Figure 4: Character appearance frequency and relative presence per episode in the dataset, as detected by the YOLOv5-based character detector. The plots illustrate how main characters maintain consistent prominence across episodes, while supporting characters show more sporadic and limited appearances, reflecting their narrative significance and posing challenges for detection model training under class imbalance.
• Figure 5: Character co-occurrence matrix illustrating the frequency with which pairs of characters appear together within the same frame across the analyzed video episodes.