Object detection based on colour of objects and length

[yeongnamtan]

A customer came to be with an interesting problem. He operates a seafood wholesaler. Today he has to assign one guy to screen all his prawns to detect if they are good enough for sale. This is done visually by inspecting the colour of the prawns.

Is there a way to differentiate object based on colour ? For example in this case, if a prawn is of a certain shade, then it is classified as not suitable for sale.

Another requirement is to know the length of the prawns and to pick out outliers eg. Prawns more than a certain length need to be set aside. But prawns are curled in general so how do we calculate its length?

[UltralyticsAssistant]

👋 Hello @yeongnamtan, thank you for sharing this interesting use case with Ultralytics! 🦐🚀 This is an automated response to help you get started while an Ultralytics engineer reviews your discussion.

We recommend checking out the Docs for comprehensive guidance on using Ultralytics models, including Python and CLI examples. Many common questions and best practices are covered there.

If this is a 🐛 Bug Report, please provide a minimum reproducible example to help us debug your issue.

If your question relates to custom training or dataset labeling, please share as much detail as possible—such as annotated image examples, training logs, and ensure you are reviewing our Tips for Best Training Results.

You can also join our community for real-time chat on Discord 🎧, participate in deeper discussions on Discourse, or connect with others on our Subreddit.

Upgrade

Make sure you are using the latest ultralytics package and all requirements in a Python>=3.8 environment with PyTorch>=1.8 to ensure the best experience and compatibility:

pip install -U ultralytics

Environments

YOLO can be run in any of these up-to-date, verified environments (with dependencies like CUDA/CUDNN, Python, and PyTorch preinstalled):

• Notebooks with free GPU:
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Amazon Deep Learning AMI. See AWS Quickstart Guide
• Docker Image. See Docker Quickstart Guide

Status

If this badge is green, all Ultralytics CI tests are currently passing. CI tests verify correct operation of all YOLO Modes and Tasks across macOS, Windows, and Ubuntu every 24 hours and on every commit.

An Ultralytics engineer will be with you soon to assist further!

[glenn-jocher]

Great use case! For color-based prawn quality assessment, you can combine YOLO11 object detection with color detection techniques. YOLO11 can detect and locate the prawns, then you can analyze the color properties within each detected region using HSV color space, which is more robust to lighting variations than RGB and better suited for color-based classification tasks.

For measuring curved prawn length, you could use YOLO11 segmentation to get the precise outline of each prawn, then calculate the arc length along the segmentation mask's contour using techniques like OpenCV's cv2.arcLength() function, which measures the perimeter of the curved shape rather than just the straight-line distance.

This approach would give you both automated quality control based on color analysis and size measurements for your seafood processing workflow.

[yeongnamtan]

Thank you for your advice. Could we also approach this by having 2 classes ... One healthy and one unhealthy and train the models based on this? Would this approach be more robust?

[glenn-jocher]

Absolutely! Training a binary classification model with "healthy" and "unhealthy" classes would likely be more robust than manual color analysis. This approach allows the model to learn complex visual patterns beyond just color - including texture, shape variations, and subtle quality indicators that humans naturally recognize but are difficult to program manually.

You can use YOLO11 for image classification by training on your labeled prawn dataset. This end-to-end approach would be more resilient to lighting changes and could capture nuanced quality features that a rule-based color detection system might miss. You'd just need to collect and annotate examples of both healthy and unhealthy prawns for training.

[yeongnamtan]

thank you for your advice.

[glenn-jocher]

You're welcome! Feel free to reach out if you run into any questions while implementing your prawn quality assessment system - it's a really interesting application of computer vision for seafood processing.