YoloEase

is a tool that automates Yolo 8+ training process. It can leverage annotation capabilities of CVAT (any server will work, by default it uses unofficial instance but you can change it at any time). Wiki is located here - https://wiki.eyeauras.net/en/YoloEase/getting-started

The tool is currently in early alpha stage. Any feedback is highly appreciated {.is-warning}

Let's Get Started

1. Install the necessary prerequisites.
2. Download latest version here
3. Setup CVAT and YoloEase projects. These will be used together for different parts of the process.
4. Dive into training using the automatic trainer.
5. Deploy and utilize your trained model!

First you setup CVAT/Dataset/model settings

Then you can start creating annotation tasks. When new task is created you can pick an option to pre-annotate the batch and/or pick only those images which will benefit model the most

As soon as the program will detect that something has changed - settings, annotations, images - it will re-train the model right away. As soon as the model will be ready you can use it to pre-annotate next batch

Two operation modes - Local Training and Google Collab

Local Training

With Local Training you're training the model using your own hardware. Positive side - you're paying only for electricity, negative - you have to have everything setup and ready to go

Google Collab

With Google Collab you're using cloud infrastructure to do the training, meaning that the actual training code runs elsewhere. You can either use a free tier or a paid one, your choice.

How it streamlines the process

• Datasets: Select your unannotated images for training. YoloEase treats these as read-only. Images/videos are supported.
• Image Management: YoloEase automatically categorizes your images (e.g., annotated, unannotated, "broken", "outliers"), ensuring efficient use of resources and no redundant work. All you have to do is throw in more images whenever needed.
• Configuration: Define your base model, training settings, and other preferences. This setup is utilized once your data is prepped and ready.
• Annotation Cycle: Annotate a set of images using CVAT and YoloEase automaticaly will re-train the model for you as soon as it will detect that new annotations are available.