Adding Classes in pretrained models without distrubing other classes

[Taiyo0004]

I want to ask how can we add classes to a pretrained model like yolov10 pretrained model so i could get the accuracy of those pretrained model and also add some of mine, is it even possible or not.

[UltralyticsAssistant]

👋 Hello @Taiyo0004, thanks for opening this discussion on adding classes to a pretrained model 🚀 This is an automated response to help you get started, and an Ultralytics engineer will also assist you soon.

We recommend a visit to the Docs for new users where you can find many Python and CLI usage examples and where many of the most common questions may already be answered.

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

If this is a custom training ❓ Question (like adding new classes while keeping existing ones), please share as much detail as you can so we can point you in the right direction, including:
-Your current ultralytics version
-Which pretrained weights you’re starting from (e.g., YOLOv10 variant)
-Your dataset YAML (classes/names) and a few example images (or a clear description)
-Your training command/config and relevant logs

Also please verify you are following our Tips for Best Training Results ✅

Join the Ultralytics community where it suits you best. For real-time chat, head to Discord 🎧. Prefer in-depth discussions? Check out Discourse. Or dive into threads on our Subreddit to share knowledge with the community.

Upgrade

Upgrade to the latest ultralytics package including all requirements in a Python>=3.8 environment with PyTorch>=1.8 to verify your issue is not already resolved in the latest version:

pip install -U ultralytics

Environments

YOLO may be run in any of the following up-to-date verified environments (with all dependencies including 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 on macOS, Windows, and Ubuntu every 24 hours and on every commit.

[glenn-jocher]

For standard closed-vocabulary detectors like yolov10*.pt, you can’t “append” classes while keeping the original head/weights unchanged—changing nc replaces the detection head, so preserving old-class accuracy requires fine-tuning on a dataset that includes both your new classes and representative data for the old ones (otherwise the old classes will typically degrade).

yolo train model=yolov10s.pt data=your_data.yaml imgsz=640 epochs=50

If your goal is adding new class names without retraining, use an open-vocabulary model like YOLO-World (prompted via set_classes())—accuracy will be scenario-dependent, but it’s the intended “add classes at runtime” path.