# How to Retrain YOLO11 on Raspberry Pi with a Custom Dataset **URL:** \ **Category:** Discussion\ **Tags:** discussion, support\ **Created:** [December 11, 2025, 11:56am UTC](https://community.ultralytics.com/t/how-to-retrain-yolo11-on-raspberry-pi-with-a-custom-dataset/1693 "2025-12-11T11:56:54Z")\ **Posts on this page:** 2\ **Page:** 1
**Author:** ![dj3000](https://avatars.discourse-cdn.com/v4/letter/d/858c86/32.png) [@dj3000](https://community.ultralytics.com/u/dj3000)\ **Post date:** [December 11, 2025, 11:56am UTC](https://community.ultralytics.com/t/how-to-retrain-yolo11-on-raspberry-pi-with-a-custom-dataset/1693/1 "2025-12-11T11:56:55Z")
Hi everyone, I’ve been experimenting with YOLO11 on a Raspberry Pi with a camera, and it seems to be working well so far. I’d now like to create my own custom dataset and retrain the existing model so that it includes all the original pre-trained data _plus_ new objects I want the system to detect. Could someone advise me on the following? 1. **Where can I find good-quality, free, open-source datasets?** 2. **How should I prepare and label the data correctly?** 3. **How can I use this new data to retrain or update an existing YOLO model?** If there are any recommended guides, tutorials, or YouTube videos that walk through this process, it would be greatly appreciated. Thanks in advance! ---
**Author:** ![BurhanQ](https://sea1.discourse-cdn.com/flex001/user_avatar/community.ultralytics.com/burhanq/32/7_2.png) [@BurhanQ](https://community.ultralytics.com/u/BurhanQ)\ **Post date:** [December 11, 2025, 2:56pm UTC](https://community.ultralytics.com/t/how-to-retrain-yolo11-on-raspberry-pi-with-a-custom-dataset/1693/2 "2025-12-11T14:56:19Z")
> [@dj3000](#): > > - **Where can I find good-quality, free, open-source datasets?** Several more options than these: - [https://datasetsearch.research.google.com](https://datasetsearch.research.google.com) - [Datasets – Hugging Face](https://huggingface.co/datasets) - [https://kaggle.com](https://kaggle.com) You can also take a look at the [datasets listed in the Ultralytics documentation](https://docs.ultralytics.com/datasets/) and check their respective terms of use. > [@dj3000](#): > > - **How should I prepare and label the data correctly?** See the Docs page on Training: > **[Model Training with Ultralytics YOLO](https://docs.ultralytics.com/modes/train/)** > > Learn how to efficiently train object detection models using YOLO11 with comprehensive instructions on settings, augmentation, and hardware utilization. Quick workflow summary: - Update data YAML with old + new labels - Annotate new images with same labels in data YAML (use whatever labeling software you like most) - Split dataset into train/validation/test (at least train/validation, but test set is useful for “true” performance testing) - Train, validate, test - Annotate more as required to improve performance There’s an updated [guide about model training](https://docs.ultralytics.com/guides/model-training-tips/) on the doc site, and this older [guide](https://docs.ultralytics.com/yolov5/tutorials/tips_for_best_training_results/), is still very relevant. Absolutely recommend reading it, as well as the articles linked in the guide. There is also [this guide on data labeling](https://docs.ultralytics.com/guides/data-collection-and-annotation/) that would also be a good read. > [@dj3000](#): > > - **How can I use this new data to retrain or update an existing YOLO model?** Download the full COCO dataset, add your new images + labels, train the model (don’t do this on the Raspberry Pi thought, just wanted ensure that was obvious). You’ll want a computer with a good CPU and a GPU with as much VRAM as possible. If you don’t own one, there are cloud based ones, but for large training you’ll likely need to pay to avoid disruptions. Training the COCO dataset for YOLO11n might take a day or so with a reasonably large GPU, could take longer with a smaller one and/or significantly more data/labels.