Description:

This dataset consists of a diverse collection of images featuring Paimon, a popular character from the game Genshin Impact. The images have been sourced from in-game gameplay footage and capture Paimon from various angles and in different sizes (scales), making the dataset suitable for training YOLO object detection models.

The dataset provides a comprehensive view of Paimon in different lighting conditions, game environments, and positions, ensuring the model can generalize well to similar characters or object detection tasks. While most annotations are accurately labeled, a small number of annotations may include minor inaccuracies due to manual labeling errors. This is ideal for researchers and developers working on character recognition, object detection in gaming environments, or other AI vision tasks.

Download Dataset

Dataset Features:

Image Format: .jpg files in 640×320 resolution.

Annotation Format: .txt files in YOLO format, containing bounding box data with:

class_id

x_center

y_center

width

height

Use Cases:

Character Detection in Games: Train YOLO models to detect and identify in-game characters or NPCs.

Gaming Analytics: Improve recognition of specific game elements for AI-powered game analytics tools.

Research: Contribute to academic research focused on object detection or computer vision in animated and gaming environments.

Data Structure:

Images: High-quality .jpg images captured from multiple perspectives, ensuring robust model training across various orientations and lighting scenarios.

Annotations: Each image has an associated .txt file that follows the YOLO format. The annotations are structured to include class identification, object location (center coordinates), and bounding box dimensions.

Key Advantages:

Varied Angles and Scales: The dataset includes Paimon from multiple perspectives, aiding in creating more versatile and adaptable object detection models.

Real-World Scenario: Extracted from actual gameplay footage, the dataset simulates real-world detection challenges such as varying backgrounds, motion blur, and changing character scales.

Training Ready: Suitable for training YOLO models and other deep learning frameworks that require object detection capabilities.

This dataset is sourced from Kaggle.