mlbb-ai-clipper

YOLO Custom Model for Object Detection

This repository contains a specially trained YOLO model designed for the epic moment UI detection task in the mobile legend (MLBB) video game. The model was trained using a public dataset without including personal information. The pre-trained model is YOLOv8m using a 6 Core 12 thread CPU device and NVIDIA RTX 3090 24GB GDDR6X.

Model Details

• Architecture: YOLO
• Framework: PyTorch
• Performance: mAP@50, precision, recall, etc.
• Model File: best.pt

---

Getting Started

Installation

Ensure you have the following dependencies installed:

pip install ultralytics

Usage

To use the model for inference, follow these steps:

from ultralytics import YOLO

# Load the model
model = YOLO("best.pt")

# Perform inference on an image
results = model("path/to/your/image.jpg")

# Display or save results
results.show()  # Show predictions
results.save("output/")  # Save predictions to output folder

Input Format

• Image: Any standard image format (e.g., JPG, PNG).
• Video: Supported for object tracking and frame-by-frame inference.

Output

The model outputs bounding boxes, class labels, and confidence scores for each detected object.

---

Example Results

Here are some example results generated by the model:

---

Training Details

• Pretrained Weights: YOLO pretrained weights were used as the starting point.
• Hyperparameters: Custom hyperparameters for learning rate, epochs, and batch size were used. Refer to the config.yaml file for details.
• Hardware: Training was conducted on NVIDIA RTX 3090.

---

Demo

You can interact with this model through a hosted Gradio app:
Gradio Demo (if available).

---

Limitations

• The model is optimized for the specific dataset it was trained on. It may not generalize well to other datasets.
• Objects not present in the training dataset may not be detected accurately.

---

License

This model is licensed under the MIT License. Feel free to use it for personal or commercial purposes with attribution.

---

Acknowledgments

• Built using Ultralytics YOLO.
• The dataset used for model training is taken from public datasets such as public streams or clips without including any personal information.

---

Contact

For questions, issues, or suggestions, feel free to reach out:

• Author: Frendy Rachman (https://huggingface.co/frendyrachman)
• Email: [email protected]

---