aqillakhamis/cat-dog-beginner-classifier

CatDog Beginner Classifier

CatDog Beginner Classifier is an image-classification model that predicts whether an input image contains a domestic cat or a domestic dog.

The model was created for educational demonstrations and beginner AI projects.

Model Details

• Model name: CatDog Beginner Classifier
• Model version: 1.0
• Model type: Image classification
• Architecture: Convolutional Neural Network
• Framework: TensorFlow and Keras
• Developed by: Replace with your name or group name
• Release date: July 2026

Model Inputs and Outputs

Input

• RGB image
• Image size: 224 × 224 pixels
• Supported formats: JPG, JPEG and PNG
• Pixel values after normalization: 0–1

Output

The model returns a probability for each class:

• Cat
• Dog

Training Data

• Total approved images: 2,000
• Cat images: 1,000
• Dog images: 1,000
• Dataset version: 1.0
• Annotation guideline version: 1.0

Dataset Split

Split	Images	Percentage
Training	1,400	70%
Validation	300	15%
Testing	300	15%

Training Procedure

Preprocessing

1. Convert images to RGB.
2. Resize images to 224 × 224 pixels.
3. Normalize pixel values to 0–1.
4. Remove corrupted images.

Hyperparameters

Parameter	Value
Epochs	20
Batch size	32
Optimizer	Adam
Learning rate	0.001
Loss function	Binary cross-entropy

Evaluation

The model was evaluated using 300 test images that were not used during model training.

Metric	Result
Accuracy	91.0%
Precision	90.0%
Recall	89.0%
F1-score	89.5%

Workshop note: Replace these example values with the actual results from your model.

Intended Use

The model may be used for:

• Beginner image-classification exercises
• Teaching AI model evaluation
• Demonstrating Model Cards
• Educational applications

Intended Users

• Students
• Lecturers
• Beginner AI developers

Out-of-Scope Use

The model should not be used for:

• Veterinary diagnosis
• Animal-health decisions
• Animal-breed identification
• Wildlife classification
• Safety-critical decisions
• Commercial applications without further evaluation

Limitations

The model may produce incorrect predictions when:

• Images are dark or blurry.
• The animal is partly hidden.
• The animal is very small.
• Multiple animals appear in the image.
• The image contains a cartoon or toy.
• The input differs significantly from the training data.

Bias and Risks

The dataset may overrepresent daylight images, common animal breeds, indoor environments and certain image sources.

A high confidence score does not guarantee that a prediction is correct.

Deployment Recommendations

1. Apply the documented preprocessing.
2. Display prediction probabilities.
3. Mark low-confidence predictions as uncertain.
4. Allow human review for important decisions.
5. Record incorrect predictions.
6. Monitor model performance.
7. Retrain the model when the data changes.

Maintenance

• Model owner: Replace with your name or group name
• Review frequency: Every six months
• Retraining trigger: Significant performance decline

Version History

Version	Date	Description
1.0	July 2026	Initial model release