---
license: apache-2.0
language:
- en
metrics:
- accuracy
base_model:
- huawei-noah/TinyBERT_General_4L_312D
tags:
- TaxonomyOfVideoGameBugs
---
# Invalid Graphical Representation Bug Detection Model

This model is trained to detect **Invalid Graphical Representation** issues in video games. It identifies bugs where aspects of the world state are incorrectly rendered. For example, it can detect when a character is performing a swimming animation while on land, or when a clothing item is not appearing as intended in the game.

The model is based on concepts from the paper [_"What went wrong; the taxonomy of video game bugs"_](https://users.soe.ucsc.edu/~ejw/papers/lewis-taxonomy-fdg2010.pdf).

## Model Details

- **Model Type**: Classification (Binary)
- **Training Data**: Trained on a dataset of video game bugs, specifically focused on **Invalid Graphical Representation** as described in the paper _"What went wrong; the taxonomy of video game bugs"_. 
- **Task**: Bug detection in video games.
- **Intended Use**: This model is designed for game developers and QA teams to automate the detection of rendering issues or animation bugs in video games.

## Training Metrics

The model was trained for 4 epochs, with the following performance metrics:

| Epoch | Training Loss | Validation Loss | Accuracy | F1 Score | Precision | Recall  |
|-------|---------------|-----------------|----------|----------|-----------|---------|
| 1     | 0.6919        | 0.6906          | 61.75%   | 0.7086   | 0.5688    | 0.9394  |
| 2     | 0.6707        | 0.6646          | 67.25%   | 0.6289   | 0.7161    | 0.5606  |
| 3     | 0.6181        | 0.5799          | 76.50%   | 0.7939   | 0.7016    | 0.9141  |
| 4     | 0.4611        | 0.4110          | 87.00%   | 0.8791   | 0.8147    | 0.9545  |

### Key Metrics:
- **F1 Score**: Balances precision and recall, with a final value of 0.8791 after 4 epochs.
- **Precision**: The accuracy of positive predictions.
- **Recall**: The ability to detect true positives.

## Intended Audience

- **Game Developers**: To detect graphical and animation bugs in video games.
- **Quality Assurance (QA) Teams**: To automate the detection of rendering or animation issues during game testing.
- **Researchers**: Interested in analyzing or extending bug detection models for video games.


## Limitations:
- This model is specifically trained to detect Invalid Graphical Representation bugs, focusing on issues like animation errors and missing rendered items.
- It may not generalize well to other types of video game bugs outside of this category.
- Performance can vary depending on the game context and rendering engine. Further fine-tuning may be required for use in different games.

## How to Use

You can use the model for **binary classification** to predict whether a given game state exhibits an **Invalid Graphical Representation** bug. Here's an example using the Hugging Face `transformers` library:

```python
from transformers import pipeline

# Load the model from Hugging Face
bug_detection = pipeline('text-classification', model='fyp-buglens/VideoGameReviews-InvalidGraphicalRepresentation-TinyBERT')

# Example usage
result = bug_detection("A character is swimming on land")
print(result)  # Output: label indicating if it's a bug or not


---
license: apache-2.0
---