I am not sure if there is any difference between retraining and creating your own custom model .But this is what i did
## Using YOLOv8 Retraining Docker
In this example, we’re going to retrain the model to detect barcodes, using the barcode-detector dataset from Kaggle. After the retraining process, we’re going to convert the model to HEF and test it on the Raspberry Pi 5 AI Kit.
### This tutorial was made on a development machine with this spec
**Hardware**:
- CPU: Intel i7-6850K
- GPU: RTX 4080
**Software**:
- OS: Ubuntu 20.04
- Hailo DFC version: 3.27.0
- Hailo Model-Zoo: 2.11.0
### On the development machine
1. Install the Hailo AI SW-Suite from the [Developer Zone](https://hailo.ai/developer-zone/software-downloads/). Alternatively, you can download and install the DFC and the model-zoo into the same virtual environment.
2. Follow the instructions on the YOLOv8 retraining page: [YOLOv8 Retraining](https://github.com/hailo-ai/hailo_model_zoo/tree/833ae6175c06dbd6c3fc8faeb23659c9efaa2dbe/training/yolov8)
3. Note in this example we added volume mount with the name `data` to the Docker container.
show original
The important point to note is you will need to use the docker in creating the pt model . Never worked for me if i created a pt model without using the docker.
I can confirm, this solved the same issue I had with a yolov8s model that I trained with the Hailo Model Zoo retraining docker.
Just for information, my model had 30 classes and was trained with 1000 images over 100 epochs.