nncase is a neural network compiler for AI accelerators.

Telegram: nncase community Technical Discussion QQ Group: 790699378 . Answer: 人工智能

K230

Install

Linux：
```
pip install nncase nncase-kpu
```

Windows:

1. pip install nncase
2. Download `nncase_kpu-2.x.x-py2.py3-none-win_amd64.whl` in below link.
3. pip install nncase_kpu-2.x.x-py2.py3-none-win_amd64.whl

All version of nncase and nncase-kpu in Release.

Supported operators

benchmark test

<table> <tr> <th>kind</th> <th> model </th><th> shape </th><th> quant_type(If/W) </th><th> nncase_fps </th><th> tflite_onnx_result </th><th> accuracy </th><th> info </th></tr> <tr> <td rowspan='3'>Image Classification</td> <td>mobilenetv2 </td><td> [1,224,224,3] </td><td> u8/u8 </td><td> 600.24 </td><td> top-1 = 71.3% top-5 = 90.1% </td><td> top-1 = 71.1% top-5 = 90.0% </td><td> dataset(ImageNet 2012, 50000 images) tflite </td></tr> <tr><td>resnet50V2 </td><td> [1,3,224,224] </td><td> u8/u8 </td><td> 86.17 </td><td> top-1 = 75.44% top-5 = 92.56% </td><td> top-1 = 75.11% top-5 = 92.36% </td><td> dataset(ImageNet 2012, 50000 images) onnx</td></tr> <tr><td>yolov8s_cls </td><td> [1,3,224,224] </td><td> u8/u8 </td><td> 130.497 </td><td> top-1 = 72.2% top-5 = 90.9% </td><td> top-1 = 72.2% top-5 = 90.8% </td><td> dataset(ImageNet 2012, 50000 images) yolov8s_cls(v8.0.207)</td></tr> <tr> <td rowspan='2'>Object Detection</td> <td>yolov5s_det </td><td> [1,3,640,640] </td><td> u8/u8 </td><td> 23.645 </td><td> bbox mAP50-90 = 0.374 mAP50 = 0.567 </td><td> bbox mAP50-90 = 0.369 mAP50 = 0.566</td><td>dataset(coco val2017, 5000 images) yolov5s_det(v7.0 tag, rect=False, conf=0.001, iou=0.65)</td></tr> <tr><td>yolov8s_det </td><td> [1,3,640,640] </td><td> u8/u8 </td><td> 9.373 </td><td> bbox mAP50-90 = 0.446 mAP50 = 0.612 mAP75 = 0.484 </td><td> bbox mAP50-90 = 0.404 mAP50 = 0.593 mAP75 = 0.45</td><td>dataset(coco val2017, 5000 images) yolov8s_det(v8.0.207, rect = False)</td></tr> <tr> <td rowspan='1'>Image Segmentation</td> <td>yolov8s_seg </td><td> [1,3,640,640] </td><td> u8/u8 </td><td> 7.845 </td><td> bbox mAP50-90 = 0.444 mAP50 = 0.606 mAP75 = 0.484 segm mAP50-90 = 0.371 mAP50 = 0.578 mAP75 = 0.396 </td><td> bbox mAP50-90 = 0.444 mAP50 = 0.606 mAP75 = 0.484 segm mAP50-90 = 0.371 mAP50 = 0.579 mAP75 = 0.397</td><td> dataset(coco val2017, 5000 images) yolov8s_seg(v8.0.207, rect = False, conf_thres = 0.0008)</td></tr> <tr> <td rowspan='3'>Pose Estimation</td> <td>yolov8n_pose_320 </td><td> [1,3,320,320] </td><td> u8/u8 </td><td> 36.066 </td><td> bbox mAP50-90 = 0.6 mAP50 = 0.843 mAP75 = 0.654 keypoints mAP50-90 = 0.358 mAP50 = 0.646 mAP75 = 0.353 </td><td> bbox mAP50-90 = 0.6 mAP50 = 0.841 mAP75 = 0.656 keypoints mAP50-90 = 0.359 mAP50 = 0.648 mAP75 = 0.357 </td><td> dataset(coco val2017, 2346 images) yolov8n_pose(v8.0.207, rect = False)</td></tr> <tr><td>yolov8n_pose_640 </td><td> [1,3,640,640] </td><td> u8/u8 </td><td> 10.88 </td><td> bbox mAP50-90 = 0.694 mAP50 = 0.909 mAP75 = 0.776 keypoints mAP50-90 = 0.509 mAP50 = 0.798 mAP75 = 0.544 </td><td> bbox mAP50-90 = 0.694 mAP50 = 0.909 mAP75 = 0.777 keypoints mAP50-90 = 0.508 mAP50 = 0.798 mAP75 = 0.54 </td><td> dataset(coco val2017, 2346 images) yolov8n_pose(v8.0.207, rect = False)</td></tr> <tr><td>yolov8s_pose </td><td> [1,3,640,640] </td><td> u8/u8 </td><td> 5.568 </td><td> bbox mAP50-90 = 0.733 mAP50 = 0.925 mAP75 = 0.818 keypoints mAP50-90 = 0.605 mAP50 = 0.857 mAP75 = 0.666 </td><td> bbox mAP50-90 = 0.734 mAP50 = 0.925 mAP75 = 0.819 keypoints mAP50-90 = 0.604 mAP50 = 0.859 mAP75 = 0.669</td><td> dataset(coco val2017, 2346 images) yolov8s_pose(v8.0.207, rect = False)</td></tr> </table>

Demo

|eye gaze | space_resize | face pose| | --- | --- | --- | | <img src="https://github.com/kendryte/nncase_docs/blob/master/gif/eye_gaze_result.gif?raw=true" alt="gif"> | <img src="https://github.com/kendryte/nncase_docs/blob/master/gif/space_resize.gif?raw=true" alt="gif">| <img src="https://github.com/kendryte/nncase_docs/blob/master/gif/face_pose_result.gif?raw=true"> |

K210/K510

Supported operators

Features

Supports multiple inputs and outputs and multi-branch structure
Static memory allocation, no heap memory acquired
Operators fusion and optimizations
Support float and quantized uint8 inference
Support post quantization from float model with calibration dataset
Flat model with zero copy loading

Architecture

Build from source

It is recommended to install nncase directly through pip. At present, the source code related to k510 and K230 chips is not open source, so it is not possible to use nncase-K510 and nncase-kpu (K230) directly by compiling source code.

If there are operators in your model that nncase does not yet support, you can request them in the issue or implement them yourself and submit the PR. Later versions will be integrated, or contact us to provide a temporary version. Here are the steps to compile nncase.

git clone https://github.com/kendryte/nncase.git
cd nncase
mkdir build && cd build

# Use Ninja
cmake .. -G Ninja -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=./install
ninja && ninja install

# Use make
cmake .. -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=./install
make && make install

Resources

Canaan developer community

Canaan developer community contains all resources related to K210, K510, and K230.

资料下载 --> Pre-compiled images available for the development boards corresponding to the three chips.
文档 --> Documents corresponding to the three chips.
模型库 --> Examples and code for industrial, security, educational and other scenarios that can be run on the K210 and K230.
模型训练 --> The model training platform for K210 and K230 supports the training of various scenarios.

Nncase

Install / Use

README

K230

Install

Supported operators

benchmark test

Demo

K210/K510

Supported operators

Features

Architecture

Build from source

Resources

Canaan developer community

Bilibili

K210 related repo

K230 related repo