Model Evaluation Terminology

Concept: A Confusion Matrix is a performance evaluation tool for classification models. It summarizes prediction results by comparing actual labels with predicted labels, helping analyze model strengths and weaknesses. Step 1: {\color{red}Structure of a Confusion Matrix}
A binary classification confusion matrix contains four components:

• True Positive (TP): Correctly predicted positive cases
• True Negative (TN): Correctly predicted negative cases
• False Positive (FP): Incorrectly predicted positive (Type I error)
• False Negative (FN): Incorrectly predicted negative (Type II error)

Step 2: {\color{red}Accuracy}
Accuracy measures overall correctness of the model: It shows the proportion of total correct predictions.
Step 3: {\color{red}Precision}
Precision measures prediction reliability for positive class: It answers: \emph{Out of predicted positives, how many were correct?}
Step 4: {\color{red}Recall (Sensitivity)}
Recall measures the model’s ability to detect actual positives: It answers: \emph{Out of actual positives, how many were correctly identified?} Step 5: {\color{red}Why These Metrics Matter}

• Accuracy works well with balanced datasets
• Precision is important when false positives are costly
• Recall is critical when missing positives is dangerous (e.g., medical diagnosis)

Concept: Computer Vision is a branch of Artificial Intelligence that allows computers to understand and interpret images and videos, similar to human vision. It enables applications such as face recognition, object detection, and medical imaging analysis. Step 1: {\color{red}Image Acquisition}
This is the first stage where visual data is captured:

• Using cameras, sensors, or imaging devices
• Sources include photos, video streams, or scanned documents

The goal is to obtain raw visual input for processing.
Step 2: {\color{red}Preprocessing}
In this stage, the acquired image is prepared for analysis:

• Noise reduction and filtering
• Image resizing or normalization
• Contrast enhancement or grayscale conversion

Preprocessing improves image quality and consistency.
Step 3: {\color{red}Feature Extraction}
This stage identifies meaningful patterns from images:

• Detecting edges, textures, or shapes
• Extracting key points or descriptors

These features are later used for tasks like classification or detection.

Concept: The Train-Test Split is a fundamental evaluation technique in machine learning used to assess a model’s ability to perform on new, unseen data. It prevents overly optimistic results by separating training and evaluation data. Step 1: {\color{red}What is Train-Test Split?}
The dataset is divided into two parts:

• Training set — used to train the model
• Testing set — used to evaluate performance

A common split is 80:20 or 70:30.
Step 2: {\color{red}Purpose of Training Data}
The training data helps:

• Learn patterns and relationships
• Adjust model parameters

The model sees this data during learning.
Step 3: {\color{red}Purpose of Testing Data}
The testing data is:

• Completely unseen during training
• Used for unbiased evaluation

It simulates real-world performance.
Step 4: {\color{red}Detecting Overfitting}
Train-test split helps identify:

• High training accuracy but low test accuracy → overfitting
• Similar performance → good generalization

Step 5: {\color{red}Ensuring Model Reliability}
This technique:

• Provides realistic performance estimates
• Prevents data leakage
• Builds trustworthy models