Deepvamja/Bug_Vision_AI
Folders and files
| Name | Name | Last commit date | ||
|---|---|---|---|---|
Β | Β | |||
Β | Β | |||
Β | Β | |||
Β | Β | |||
Β | Β | |||
Β | Β | |||
Β | Β | |||
Β | Β | |||
Repository files navigation
π BugVision AI
Intelligent Bug Screenshot Classification & Explainability System
BugVision AI is a production-style deep learning system that automatically classifies software bug screenshots into real-world error categories and explains why the model made each decision using visual heatmaps.
π The Problem (Real & Industrial):
In modern software teams:
QA engineers report bugs using screenshots
Developers manually read and triage bugs
Large teams receive hundreds of screenshots daily
Bug categorization becomes slow, error-prone, and inconsistent
β Manual triaging
β Delayed debugging
β No visibility into recurring bug patterns
BugVision AI solves this by automating bug classification using computer vision.
π‘ The Solution
π§ What BugVision AI Does
βοΈ Accepts bug screenshots as input
βοΈ Classifies them into 6 real-world error types
βοΈ Generates confidence scores
βοΈ Explains predictions using Grad-CAM heatmaps
βοΈ Stores prediction history
βοΈ Provides an analytics dashboard
βοΈ Runs as a full-stack deployed application
π·οΈ Supported Bug Categories
Class Description
UI_Error Layout issues, misaligned buttons, broken UI elements
Database_Error SQL errors, DB connection failures
Network_Error Timeouts, connectivity issues
Rendering_Error Graphics glitches, blank screens
Crash_Error App crashes, fatal error screens
Other_Error Unclassified or rare error types
β¨ Key Features
π§ Automated Bug Triage
Automatically categorizes bug screenshots into real-world error types, eliminating manual inspection and reducing triage time.
π Explainable AI for Trust & Debugging
Generates Grad-CAM heatmaps that visually explain model decisions, helping engineers verify predictions instead of blindly trusting AI.
βοΈ Production-Ready Inference Pipeline
Implements a complete ML inference workflow including preprocessing, prediction, confidence scoring, and result rendering β mirroring real production systems.
π Engineering Analytics Dashboard
Tracks prediction history and class distributions, enabling teams to identify recurring bug patterns and systemic issues.
ποΈ Persistent Prediction Logging
Stores inference results with timestamps, supporting auditing, debugging, and future model improvements.
π§© System Architecture
User Screenshot
β
Image Preprocessing (256Γ256 RGB)
β
CNN-based Deep Learning Model
β
Softmax Probability Distribution
β
Grad-CAM Explainability
β
Flask Web Application
β
Prediction Logging + Dashboard
π οΈ Tech Stack
ML / DL: TensorFlow, CNN, Grad-CAM
Backend: Python, Flask
Computer Vision: OpenCV, NumPy
Data: SQLite
HOW TO RUN LOCALLY
------------------
1. Create virtual environment (optional but recommended)
python -m venv venv
2. Activate environment
(Windows) venv\Scripts\activate
(Mac/Linux) source venv/bin/activate
3. Install dependencies
pip install -r requirements.txt
4. Run Flask server
python app/app.py
5. Open in browser:
http://127.0.0.1:5000