Squid QC Lab
Computer Vision · Food Quality Control

AI Drying Squid
Quality Control System

ระบบควบคุมคุณภาพปลาหมึกตากแห้งด้วยปัญญาประดิษฐ์

A camera watches squid dry on the rack. An AI model reads each piece — how dry it is, its colour, and any defects — then sorts it into Grade A, B or C in real time. Press play and watch the line run.

3
quality grades
4
features read
~0.3s
per squid
Inspection line — standby
SOLAR DRYING BAY 01

Current sample  — —

·
Waiting…
confidence —

Features detected

Dryness
Colour uniformity
Shape / size
Surface defects
A
0
Export
B
0
Local market
C
0
Re-dry / reject
Σ
0
Inspected
Air temp
°C
Humidity
%
Belt speed
pcs/min
Run time
0:00
SPEED
How the AI decides

Four steps, from camera to grade

The system does the same thing an expert inspector does by eye and by touch — only faster, and the same way every single time.

STEP 01

Capture

A camera above the drying rack photographs each squid as it passes on the belt under even light.

STEP 02

Read features

The model measures four things from the image: dryness, colour uniformity, shape & size, and surface defects like mould or tears.

STEP 03

Classify

It combines the scores and predicts a grade — A, B or C — with a confidence level, in a fraction of a second.

STEP 04

Sort

The grade tells the line where each squid goes: export, local market, or back to the rack to dry more.

The rulebook

What makes a squid Grade A, B or C

These are the thresholds the model is trained on. In the simulation above, every squid is scored against them live.

GradeDrynessColourDefectsWhere it goes
A ≥ 85% Even, golden None Export grade
B 70–84% Slightly uneven Minor spots Local market
C < 70% Dark / patchy Mould, tears Re-dry or reject

Why it matters

  • Fairer pricing. Every squid is judged by the same rules, so buyers and sellers trust the grade.
  • Less waste. Squid that isn't dry enough is caught and sent back to dry, instead of spoiling in the bag.
  • Faster. One camera can check thousands of pieces a day without getting tired.
  • Safer food. Mould and defects are flagged before the product reaches a plate.

How a student could build this

  • Collect photos of squid at different dryness levels and label them A / B / C.
  • Train a model with a no-code tool like Google Teachable Machine, or a small CNN in Python.
  • Test it on new photos and check how often it agrees with a human grader.
  • Add sensors (temperature & humidity) to watch the drying environment too.
Teachable Machine Python + TensorFlow OpenCV Raspberry Pi camera DHT22 sensor