YOLOv8-RKNN2-Ultralytics

Converted with the Ultralytics export workflow and task-specific RKNN conversion tools. This repository contains 100 FP16 and INT8 exports for RK3576 and RK3588. Tasks include classify, detect, obb, pose, segment.

Choose a model under models/<toolkit-version>/<task>/<variant>/<chip>/<precision>/ and download both model.rknn and metadata.yaml. The manifest lists the supported combinations and SHA-256 hashes.

Source and license

The inputs are official Ultralytics checkpoints. Conversion changed the file format and quantized the INT8 graphs. The RKNN files and original weights are distributed under AGPL-3.0. See the license and checkpoint URLs and hashes.

Run python source/convert-model.py --help. For a detection INT8 export, pass a COCO128 dataset YAML. For other INT8 tasks, pass a COCO128 image list. Use --calibration-limit 128 and the input size in the manifest.

The source directory contains the original checkpoints, conversion scripts, and environment and calibration records. Upstream Ultralytics source is available from PyPI. The conversion used RKNN Toolkit2 2.3.2. See RKNN Toolkit2, available separately from Rockchip.

INT8 detection

Use the metadata.yaml from the same INT8 model folder. The detector outputs normalized box coordinates, so the app needs coordinates: normalized. Import both files together in the warehouse. For the CLI, keep them in the same directory. Interpreting the coordinates as pixels can result in no boxes.

Models are organized by RKNN Toolkit2 version under models/v<version>/. Each model and its metadata share a folder.