NO code YOLOv5 model with Ultralytics HUB

This article is about how to create a NO-code object detection YOLOv5 with Ultralytics HUB.

Result from Ultralytics HUB

To increase the speed of project implementation, I’ve been working on simplify the training process by create my shared notebook in my previous article which is still hard for non-coder user and only some super users can use that framework.

But things will be different because of the Ultralytics HUB.

What is Ultralytics HUB?

Ultralytics HUB is a platform that created by Ultralytics team, the same team that created YOLOv5, they launched beta version in early of 2022.

With Ultralytics HUB, we can create object detection models with no code.
*At the time I wrote this article, Ultralytics HUB is beta version v0.1.14-beta.5

Label and organize your dataset

I will use the same dataset as I used in my previous article.

Steps are similar to the previous except your folder structure,

1. Organize our files and folders like this.

-your project name (My project name is POM)
|-images
 |-train
 |-val
 |-test (optional)
|-labels
 |-train
 |-val
 |-test (optional)

2. Label your images and save label files in labels folder.
*for more info > see my previous article

Label my images

3. Create .yaml file with content inside it like this.

path : #
train: images/train
val: images/val
test: images/test
nc: n
names : ['label 1' , 'label 2' , ... , 'label n']

My .yaml file’s content and final folder structure

*It is very important to have the same folder structure as mine.

4. Zip everything.

Zip everything into a zip file

Train model(s) with Ultralytics HUB

Before begin this part, you need to open Ultralytics HUB to sign up and log in your account first. After logged in, you will see this landing page.

My landing page

After that, processes are really simple to do without any manual.

But you can follow my steps if you want, a lot of pictures here...

1. Create Project

Create New Project

2. Upload dataset

Upload dataset zip from previous step

3. Train model

Create Model and select model’s architecture

Select training dataset

Copy API key and open training instance (Colab)

Paste API key and run

Monitoring model’s loss during training

4. See model’s result

Trained model in this project

Model’s loss during the training

5. Test if needed

Test result

6. Export model(s)

Available export formats

And that’s how to create YOLOv5 model without any coding.

If you want to deploy your models, you can find YOLOv5’s link HERE or you can use my other previous article HERE.