{"id":21802457,"url":"https://github.com/qengineering/radxa-zero-3-npu-ubuntu22","last_synced_at":"2025-08-16T16:06:36.034Z","repository":{"id":235419924,"uuid":"790569909","full_name":"Qengineering/Radxa-Zero-3-NPU-Ubuntu22","owner":"Qengineering","description":"Radxa Zero 3W/E image with Ubuntu 22, OpenCV, deep learning frameworks and NPU drivers","archived":false,"fork":false,"pushed_at":"2025-06-11T08:16:48.000Z","size":34,"stargazers_count":40,"open_issues_count":3,"forks_count":3,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-06-11T09:27:54.662Z","etag":null,"topics":["ncnn","npu","opencv","radxa-zero-3e","radxa-zero-3w","radxa-zero3","radxa-zero3-npu","radxa-zero3w","rk3566","rknpu-model-zoo","rknpu2","ubuntu22"],"latest_commit_sha":null,"homepage":"https://qengineering.eu/deep-learning-examples-on-raspberry-32-64-os.html","language":null,"has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"bsd-3-clause","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/Qengineering.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2024-04-23T05:56:49.000Z","updated_at":"2025-06-11T08:16:52.000Z","dependencies_parsed_at":"2024-04-23T10:13:17.089Z","dependency_job_id":"55a90859-1015-471d-95ed-a070f04719cb","html_url":"https://github.com/Qengineering/Radxa-Zero-3-NPU-Ubuntu22","commit_stats":null,"previous_names":["qengineering/radxa-zero3-npu-ubuntu22"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/Qengineering/Radxa-Zero-3-NPU-Ubuntu22","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Qengineering%2FRadxa-Zero-3-NPU-Ubuntu22","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Qengineering%2FRadxa-Zero-3-NPU-Ubuntu22/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Qengineering%2FRadxa-Zero-3-NPU-Ubuntu22/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Qengineering%2FRadxa-Zero-3-NPU-Ubuntu22/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/Qengineering","download_url":"https://codeload.github.com/Qengineering/Radxa-Zero-3-NPU-Ubuntu22/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Qengineering%2FRadxa-Zero-3-NPU-Ubuntu22/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":270734952,"owners_count":24636271,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","status":"online","status_checked_at":"2025-08-16T02:00:11.002Z","response_time":91,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["ncnn","npu","opencv","radxa-zero-3e","radxa-zero-3w","radxa-zero3","radxa-zero3-npu","radxa-zero3w","rk3566","rknpu-model-zoo","rknpu2","ubuntu22"],"created_at":"2024-11-27T11:28:48.599Z","updated_at":"2025-08-16T16:06:36.026Z","avatar_url":"https://github.com/Qengineering.png","language":null,"funding_links":["https://www.paypal.com/cgi-bin/webscr?cmd=_s-xclick\u0026hosted_button_id=CPZTM5BB3FCYL"],"categories":[],"sub_categories":[],"readme":"# Radxa Zero 3 NPU with Ubuntu 22.04\n![output image]( https://qengineering.eu/github/RadxaZero3_GitHub.webp)\u003cbr/\u003e\u003cbr\u003e\n[![License](https://img.shields.io/badge/License-BSD%203--Clause-blue.svg)](https://opensource.org/licenses/BSD-3-Clause)\u003cbr/\u003e\n\n------------\n\n## Installation.\n\n- Get a 16 GB (minimal) SD card holding the image. \n- Download the `Radxa_Zero3_NPU_Ubuntu22.img.xz` image (2.9 GByte) from our [Sync](https://ln5.sync.com/dl/9c6592390/rsr2pb66-93y5zph6-nryj9bdt-aju7pfwf) site.\n- Flash the image on the SD card with the [Imager](https://www.raspberrypi.org/software/) or [balenaEtcher](https://www.balena.io/etcher/).\n- Insert the SD card in your Rock 5 and enjoy.\n- Username: ***radxa***\n- no password: ***radxa***\n\n------------\n\n## Model performance benchmark(FPS)\n\nAll models, with C++ examples, can be found on the SD image.\u003cbr\u003e\n\n| demo             | model_name                   | inputs_shape            | dtype | Radxa Zero3|\n| ---------------- | ---------------------------- | ----------------------- | ----- | ------------- |\n| yolov5           | yolov5s_relu                 | [1, 3, 640, 640]        | INT8  | 14.8          |\n|                  | yolov5n                      | [1, 3, 640, 640]        | INT8  | 19.5          |\n|                  | yolov5s                      | [1, 3, 640, 640]        | INT8  | 11.7          |\n|                  | yolov5m                      | [1, 3, 640, 640]        | INT8  | 5.7           |\n| yolov6           | yolov6n                      | [1, 3, 640, 640]        | INT8  | 18.0          |\n|                  | yolov6s                      | [1, 3, 640, 640]        | INT8  | 8.1           |\n|                  | yolov6m                      | [1, 3, 640, 640]        | INT8  | 4.5           |\n| yolov7           | yolov7-tiny                  | [1, 3, 640, 640]        | INT8  | 16.1          |\n|                  | yolov7                       | [1, 3, 640, 640]        | INT8  | 3.4           |\n| yolov8           | yolov8n                      | [1, 3, 640, 640]        | INT8  | 18.2          |\n|                  | yolov8s                      | [1, 3, 640, 640]        | INT8  | 8.9           |\n|                  | yolov8m                      | [1, 3, 640, 640]        | INT8  | 4.4           |\n| yolox            | yolox_s                      | [1, 3, 640, 640]        | INT8  | 10.0          |\n|                  | yolox_m                      | [1, 3, 640, 640]        | INT8  | 4.8           |\n| ppyoloe          | ppyoloe_s                    | [1, 3, 640, 640]        | INT8  | 9.2           |\n|                  | ppyoloe_m                    | [1, 3, 640, 640]        | INT8  | 5.0           |\n| yolov5_seg       | yolov5n-seg                  | [1, 3, 640, 640]        | INT8  | 1.04          |\n|                  | yolov5s-seg                  | [1, 3, 640, 640]        | INT8  | 0.87          |\n|                  | yolov5m-seg                  | [1, 3, 640, 640]        | INT8  | 0.71          |\n| yolov8_seg       | yolov8n-seg                  | [1, 3, 640, 640]        | INT8  | 0.91          |\n|                  | yolov8s-seg                  | [1, 3, 640, 640]        | INT8  | 0.87          |\n|                  | yolov8m-seg                  | [1, 3, 640, 640]        | INT8  | 0.7           |\n| RetinaFace       | RetinaFace_mobile320         | [1, 3, 320, 320]        | INT8  | 88.5          |\n|                  | RetinaFace_resnet50_320      | [1, 3, 320, 320]        | INT8  | 11.8          |\n| PPOCR-Det        | ppocrv4_det                  | [1, 3, 480, 480]        | INT8  | 15.1          |\n| PPOCR-Rec        | ppocrv4_rec                  | [1, 3, 48, 320]         | FP16  | 17.3          |\n\n* Due to the pixel-wise filling and drawing, segmentation models are relatively slow\n\n------------\n\n## Cooling.\n\nYou must cool your Zero3. It will get very hot without a heatsink.\u003cbr\u003e\nWe used a heatsink with two fans designed for the Raspberry Pi Zero, and it works fine.\u003cbr\u003e\nEven with the NPU running 24/7 at 1.8 GHz, it never gets warmer than 42°C (107°F).\u003cbr\u003e\n\n\n![output image]( https://qengineering.eu/github/RadxaZero3_Fan3.webp)\u003cbr/\u003e\u003cbr/\u003e\n\nUse the thermal pad properly.\u003cbr\u003e\nIt should fill the space between the chip and the cooling element effectively.\u003cbr\u003e\nIf there is any gap, the heat flow will not be optimal, resulting in a much hotter CPU.\u003cbr\u003e\nThe delivered pad will come with two plastic protective sheets. These must be removed before applying the pad.\u003cbr\u003e\nIf there is still a small gap (the CPU of the Radxa is slightly thinner than the Raspberry Pi), cut some small slices from the pad and stack them.\u003cbr\u003e\nThe actual CPU core is located at the centre of the chip, where the heat is generated.\u003cbr\u003e\u003cbr\u003e\n\n![output image]( https://qengineering.eu/github/RadxaZero3_FanPad.webp)\u003cbr/\u003e\u003cbr/\u003e\n\n------------\n\n## Tips.\n\n* If you need extra space delete the opencv and the opencv_contrib folder from the SD card. They are no longer needed since all libraries are stored in the /usr/ directory.\n* Use a tool like [GParted](https://gparted.org/) `sudo apt-get install gparted` to expand the image to larger SD cards. We recommend a minimum of 64 GB. Deep learning requires a lot of space.\u003cbr/\u003e\n\n\n------------\n\n## Pre-installed frameworks.\n\n- [OpenCV](https://qengineering.eu/deep-learning-with-opencv-on-raspberry-pi-4.html) 4.10.0\n- NPU [rknpu2](https://github.com/airockchip/rknn-toolkit2/tree/master/rknpu2) 1.5.2\n- NPU [model zoo](https://github.com/airockchip/rknn_model_zoo) 2.0.0\n- NPU [model zoo models](https://github.com/Qengineering/rknn_model_zoo) 2.0.0\n\n------------\n\n### Thanks.\nA more than special thanks to [***Joshua Riek***](https://github.com/Joshua-Riek) for all the hard work on the Ubuntu OS.\n\n------------\n\n[![paypal](https://qengineering.eu/images/TipJarSmall4.png)](https://www.paypal.com/cgi-bin/webscr?cmd=_s-xclick\u0026hosted_button_id=CPZTM5BB3FCYL) \u003cbr\u003e\u003cbr\u003e\n![output image]( https://qengineering.eu/github/RadxaCover.webp)\u003cbr/\u003e\u003cbr\u003e\n\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fqengineering%2Fradxa-zero-3-npu-ubuntu22","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fqengineering%2Fradxa-zero-3-npu-ubuntu22","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fqengineering%2Fradxa-zero-3-npu-ubuntu22/lists"}