kkkk

1. SRS
Software Requirements Specification
Product Defect Detection Using
Machine Learning
Guided by:
Mrs. Kanchi K Sen
Asst. Professor, CSE Department
Presented by:
Group Number 5
Mohammed Farhan Hassan K M (YCE21CS036)
Nihal Koya Thangal N (YCE21CS040)
Benita Babu (YCE21CS022)
R Rahima Afrin (YCE21CS048)
Date: April 28, 2024

2. Contents
1 Introduction 2
1.1 Purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Product Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Intended Audience and Document Overview . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 Overall Description 2
2.1 Product Perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Product Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.3 User classes and Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.4 Design and Implementation Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.5 Assumptions and Dependencies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
3 External Interface Requirements 4
3.1 User Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.2 Hardware Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.3 Software Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
4 System Features 5
4.1 Data Collection and Image Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4.2 Model Training . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4.3 Classification and Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
5 Other Nonfunctional Requirements 6
5.1 Performance Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
5.2 Safety Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
5.3 Security Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
5.4 Software Quality Attributes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
5.5 Business Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1

3. 1 Introduction
1.1 Purpose
The concerned project aims to automate the process of detecting defects and irregularities in manufac-
tured products using machine learning models.
1.2 Product Scope
In modern manufacturing processes, manual inspection of products for defects is time-consuming and
error-prone. This project addresses the challenge by automating defect detection using machine learning
techniques.
1.3 Intended Audience and Document Overview
This document provides a comprehensive overview of the software product, its parameters, and goals.
It outlines the target audience and details the user interface, as well as the software requirements for
optimal performance.
2 Overall Description
2.1 Product Perspective
Our product aims to solve the problem of detecting defects and irregularities in manufactured products.
The conventional method includes manual inspection of products for defects which is time-consuming
and prone to errors.
2.2 Product Functions
• Image Classification: The system can classify images of products as either defective or non-defective
based on features learned from training data.
• Anomaly Detection: It can detect anomalies in product images or sensor data that deviate from
normal patterns, indicating potential defects.
• Quality Control: The system can enforce quality control by automatically flagging products that
don’t meet predefined quality standards.
2.3 User classes and Characteristics
1. Manufacturing Operators
• Monitor real-time defect detection results
• Receive alerts about defective products
2. Quality Control Inspectors
• Review automated defect detection results
• Verify detected defects
2

4. 2.4 Design and Implementation Constraints
• Hardware Constraints
• Tool specification: Anaconda Navigator
• IDE: PyCharm, Jupyter Notebook, Google Colab
• Language: Python 3.0
2.5 Assumptions and Dependencies
• Assumptions:
− Sufficient and Representative Training Data
− Feature Relevance and Extractability
• Dependencies:
− Data Quality and Availability
− Feature Engineering and Selection
− Machine Learning Algorithms and Models
3

5. 3 External Interface Requirements
3.1 User Interface
• Defect Detection Results: Displays detected defective product along with relevant information
• Model Performance Metrics: Presents performance metrics for machine learning models, such as
accuracy, precision, recall, and F1 score, to assess the reliability and effectiveness of defect detection.
3.2 Hardware Interfaces
• Data Sources Integration: The system interface with data sources such as cameras to obtain raw
data for defect detection.
• Computing Hardware: High-performance computing hardware including CPUs (Central Processing
Units) and GPUs (Graphics Processing Units) is used to train machine learning models for defect
detection.
• Alerting and Notification Systems: Integration with alerting and notification systems is necessary
to inform Manufacturing Operators about detected defects or anomalies in real-time
3.3 Software Interfaces
• Preprocessing Tools Integration: External tools or software required for preprocessing raw data
before inputting it into the machine learning models to perform tasks such as data cleaning, nor-
malization, feature extraction or image enhancement.
• Model Training and Evaluation Interface: This interface provides tools and libraries for training,
evaluating, and fine-tuning machine learning models.
• Reporting Interface: This interface supports integration with reporting tools for generating reports,
graphs and analytics from defect detection process.
4

6. 4 System Features
4.1 Data Collection and Image Processing
1. Description: Our first step is to Collect Data. Gather a diverse dataset containing information
about different products and their potential defects. Next, we Preprocess the Data. Clean and
organize the collected data to prepare it for effective use by the machine learning model.
2. Functional Requirements:
• The system should have access to raw data from various sources, such as cameras.
• Jpeg Images of defective and working samples of the concerned product will be used to train
the model
• The image processor should perform the functions of resizing the images to a standard reso-
lution and flattening the image
4.2 Model Training
1. Description: Train the machine learning model using the prepared dataset, ensuring it can accu-
rately identify defects.
2. Functional Requirements:
• The dataset should be split into training and testing datasets
• The dataset the model trains on should have the same set of dependent and independent
variables
4.3 Classification and Evaluation
1. Description: The image classification is to be performed using various supervised machine learning
algorithm based models and the performance of each model is evaluated.
2. Functional Requirements:
• The classifier must provide a robust means of detecting defaults in a target product.
• Evaluation should entail displaying metrics such as accuracy, precision and F1 score
5

7. 5 Other Nonfunctional Requirements
5.1 Performance Requirements
• The system should achieve high accuracy and precision
• Response time should allow the system to keep up with the production pace
• The system must be available anytime during working hours
5.2 Safety Requirements
• The defect detection system should comply with industry regulations, standards, and safety re-
quirements
5.3 Security Requirements
• Ensure compliance with data privacy regulations and security standards
5.4 Software Quality Attributes
• Scalability
• Robustness
• Maintainability
• Adaptability
• Flexibility
5.5 Business Rules
Since the system involves the employment of free version of machine learning models, commercial viability
of our system is restricted to some extent
6