Skip to content

README.md

Latest commit

 

History

History
185 lines (127 loc) · 5.65 KB

README.md

File metadata and controls

185 lines (127 loc) · 5.65 KB

SHM Detect Tool — Shmoo Log Analysis with CNN

Automatically classify shmoo plot pass/fail patterns using a PyTorch CNN

PyPI Version Python Versions License PyPI Downloads

Table of Contents

Overview

The SHM Detect Tool uses a CNN (Convolutional Neural Network) built on PyTorch to automatically analyse shmoo plot datalogs from ATE (Automated Test Equipment). It parses raw shmoo text logs, converts them into image-like tensors, and classifies each plot's pass/fail pattern — replacing tedious manual inspection with fast, repeatable AI-based detection.

Background

  • Challenge: Engineers spend significant time visually inspecting shmoo plots to identify pass/fail patterns, margins, and anomalies across multiple test sites and conditions.
  • Solution: SHM Detect Tool automates this process by training an AlexNet-based CNN to classify shmoo patterns, generating structured XLSX reports with pass/fail highlighting and site-to-site correlation.

Features

  • CNN-Based Shmoo Classification — AlexNet architecture trained on labeled shmoo datasets
  • Shmoo Log Parsing — convert raw text shmoo logs to structured CSV with configurable keyword patterns
  • XLSX Report Generation — automated reports with pass/fail highlighting per shmoo plot
  • Site-to-Site Correlation — compare shmoo results across multiple test sites with overlay reports
  • Parallel Plot Support — side-by-side multi-site shmoo comparison in XLSX
  • CLI Interface — scriptable commands for training, analysis, and correlation
  • GUI Interface — PyQt5-based desktop application for interactive workflows
  • Web App Interface — Streamlit-based web UI for browser-based analysis
  • Configurable Parsing — JSON config files to support different shmoo log formats (S2S, CHAR Studio, etc.)

Installation

Requirements

  • Python 3.9 – 3.14
  • Windows, macOS, or Linux

Install from PyPI (Recommended)

# Basic installation (CLI + core analysis)
pip install shm-detect-tool

# With GUI support (PyQt5)
pip install shm-detect-tool[gui]

# With web app support (Streamlit)
pip install shm-detect-tool[webapp]

After installation, the shm-detect command is immediately available:

shm-detect --version
shm-detect --help

Install from Source

git clone https://github.com/showjim/SHM_Detect_Tool.git
cd SHM_Detect_Tool
python -m venv venv
venv\Scripts\activate        # Windows
# source venv/bin/activate   # macOS/Linux
pip install -e .[gui,webapp]

Usage

CLI Usage

# Analyse a shmoo log and generate an XLSX report
shm-detect analyse --file shmoo.txt --config SHM_keywords_setting.json

# Train a new CNN model
shm-detect train --dataset custom_SHM_data.csv

# Compare shmoo results across multiple log files
shm-detect correlate --files file1.txt file2.txt --config SHM_keywords_setting.json

# Show version
shm-detect -V

See doc/CLI_USAGE.md for the full CLI reference including all options and examples.

GUI Usage

Launch the desktop application:

python SHM_Detect_Tool.py

Workflow

  1. Convert Shmoo — load a raw shmoo .txt log and convert it to CSV
  2. Detect Shmoo — select the CSV file, run CNN inference, and review the generated XLSX report
  3. Train CNN — (optional) train a new model from a labeled dataset

Web App Usage

Launch the Streamlit-based web interface:

streamlit run SHM_Detect_Tool_webapp.py

JSON Configuration

The tool uses a JSON config file to define parsing rules for different shmoo log formats (CHAR Studio, customised, etc.). Each format has unique keywords for identifying site info, test items, axis boundaries, and pass/fail symbols.

Config Fields

Field Description
keyword_site Keyword that identifies site information lines in the log
keyword_item Keyword that identifies test instance or item lines
keyword_start Keyword marking the Y-axis label start (shmoo body begins after this line)
keyword_end Keyword marking the X-axis end boundary
keyword_pass Regex pattern matching pass symbols (e.g., P, *, +)
keyword_fail Regex pattern matching fail symbols (e.g., ., #, -, E)
keyword_y_axis_pos Y-axis label position: "left" (customised) or "right" (CHAR Studio)

The package ships with a default CHAR Studio config:

{
  "keyword_site": "Site",
  "keyword_item": "<",
  "keyword_start": "Tcoef(AC Spec)",
  "keyword_end": "X Axis: Vcoef(DC Spec)",
  "keyword_pass": "\\+",
  "keyword_fail": "-|E",
  "keyword_y_axis_pos": "right"
}

See doc/CLI_USAGE.md for complete examples, regex notes, and format-specific details.

Screenshots

SHM Detect Tool

License

This project is licensed under the GNU General Public License v3.0.

Built for the semiconductor ATE community