Meta RKNN Toolkit
Project description
metarknn
metarknn部署通用框架
1、安装最新版 meta-cv
pip install meta-cv
2、安装最新版 meta-rknn
pip install meta-rknn
3、目标检测示例(参考detection_demo.py代码)
import platform, cv2
import metarknn as mr
Detection = mr.DetectionRKNN
y = Detection(model_path='models/yolov8m.rknn',
input_width=640,
input_height=480,
confidence_thresh=0.5,
nms_thresh=0.3,
class_names=classnames,
device_id=0)
# 如需本地运行,需调用下面一句进行模型转换并加载,板端无需运行
if platform.machine() == 'x86_64':
y.convert_and_load(quantize=False, # 是否int8量化
dataset='dataset.txt', # 量化使用图片路径文件
is_hybrid=True) # 是否进行混合量化
img = cv2.imread('models/bus.jpg')
_dets, _scores, _labels = y.predict(cv2.cvtColor(img, cv2.COLOR_BGR2RGB), use_preprocess=True)
# 显示
y.show(img, _dets, _scores, _labels)
cv2.imwrite("models/bus.png", img)
4、实例分割示例(参考segment_demo.py代码)
import platform, cv2
import metarknn as mr
Segment = mr.SegmentRKNN
y = Segment(model_path='models/yolov8m-seg.rknn',
input_width=640,
input_height=480,
confidence_thresh=0.5,
nms_thresh=0.3,
class_names=classnames,
device_id=0)
# 如需本地运行,需调用下面一句进行模型转换并加载,板端无需运行
if platform.machine() == 'x86_64':
y.convert_and_load(quantize=False, # 是否int8量化
dataset='dataset.txt', # 量化使用图片路径文件
is_hybrid=True) # 是否进行混合量化
img = cv2.imread('models/bus.jpg')
_dets, _scores, _labels = y.predict(cv2.cvtColor(img, cv2.COLOR_BGR2RGB), use_preprocess=True)
# 显示
y.show(img, _dets, _scores, _labels)
cv2.imwrite("models/bus.png", img)
5、模型转换与量化(本地运行)
import metarknn as mr
from mr.quantization import Quantization
q = Quantization(model_path, # onnx模型路径
dataset, # dataset文件路径
output_names=["output0", "output1"]) # 定义模型输出层
if is_hybrid: # 是否混合量化
self.model = q.hybrid_convert()
else: # 非混合量化(是否int8量化)
self.model = q.convert(quantize)
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
meta-rknn-0.1.0.tar.gz
(4.7 kB
view details)
Built Distribution
meta_rknn-0.1.0-py3-none-any.whl
(10.2 kB
view details)
File details
Details for the file meta-rknn-0.1.0.tar.gz
.
File metadata
- Download URL: meta-rknn-0.1.0.tar.gz
- Upload date:
- Size: 4.7 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/3.8.0 pkginfo/1.9.6 readme-renderer/34.0 requests/2.27.1 requests-toolbelt/1.0.0 urllib3/1.26.15 tqdm/4.64.0 importlib-metadata/4.2.0 keyring/23.4.1 rfc3986/1.5.0 colorama/0.4.5 CPython/3.6.13
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 26a26e86a89de83e6039649d11d42dff3eb174c6e154e1ef257d2edf71c44159 |
|
MD5 | 3532a8630651d64cdaff7caa98123442 |
|
BLAKE2b-256 | eb729e8105f3110be551145d9808c51a81d91160c3168610029d408be63ecb1b |
File details
Details for the file meta_rknn-0.1.0-py3-none-any.whl
.
File metadata
- Download URL: meta_rknn-0.1.0-py3-none-any.whl
- Upload date:
- Size: 10.2 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/3.8.0 pkginfo/1.9.6 readme-renderer/34.0 requests/2.27.1 requests-toolbelt/1.0.0 urllib3/1.26.15 tqdm/4.64.0 importlib-metadata/4.2.0 keyring/23.4.1 rfc3986/1.5.0 colorama/0.4.5 CPython/3.6.13
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | ae17dbedbc6153301425e774783bcf4bc297774c61c0709e768a965f319a1bfc |
|
MD5 | 8a65c3e6e021651c52e1a6799ba3eb72 |
|
BLAKE2b-256 | 3c5caf4565c950c342289a08180ba4ea55acfddac011776950d9d7d73c609bed |