Github

點我去看程式碼

Motivation

In pharmaceutical manufacturing, products often suffer minor damage or defects under varying environmental conditions. For this hackathon project, we set out to address that challenge by leveraging Advantech’s ICAM and AWS cloud services to build an automated inspection and reporting system. Our goal is to continuously detect issues on the production line, automatically notify stakeholders, and generate AI-driven reports so managers can clearly understand the status of each line and take actions to reduce breakage rates.

Features

• Continuous edge monitoring with ICAM:
  ICAM captures images from the production line in real time and performs on-device (edge) inference to identify defects immediately, reducing latency and network dependence.

• Admin Dashboard for operational visibility:
  A central dashboard visualizes key metrics across production lines, such as throughput, defect rates, and drug categories, enabling data-driven decisions and faster incident response.

• AWS-powered backend and data pipeline:
  AWS cloud services provide a scalable backend for data storage and processing, orchestrating the end-to-end pipeline from edge events to analytics, notifications, and AI-generated reports.

System Workflow

1. Edge analysis on ICAM
   ICAM devices run lightweight computation close to the camera feed, performing real-time checks on captured images to assess product condition before data leaves the factory floor.

2. Event ingestion to the AWS backend
   The results of ICAM’s on-device analysis—such as defect flags, timestamps, and contextual metadata—are transmitted to the AWS backend for reliable ingestion and further processing.

3. Processing and frontend updates
   The backend aggregates and processes incoming data, then updates the frontend dashboard with live insights, including per-line throughput, defect rates, and product categorization.

4. Automated detection-to-AI reporting
   When defect rates exceed defined thresholds, the backend automatically triggers AI-driven report generation, producing detailed summaries and root-cause hypotheses for managerial review.

5. Operational feedback and parameter tuning
   Based on AI reports and dashboard insights, managers can adjust environmental parameters—such as temperature and humidity—to proactively reduce product damage and improve yield.