用opencv部署nanodet目标检测，包含C++和Python两种版本程序的实现
用opencv部署nanodet目标检测，包含C++和python两种版本程序的实现， 使用opencv里的dnn模块加载网络模型，图像预处理和后处理模块是使用C++和python编程实现。 整个程序运行，不依赖任何深度学习框架， 在windows系统和ubuntu系统，在cpu和gpu机器上都能运行。
python版本的主程序是main_nanodet.py， c++版本的主程序是main.cpp
程序里提供输入图片尺寸320和416这两种选择，类别置信度阈值confThreshold，nms重叠率阈值nmsThreshold可自行调整
nanodet的后处理模块的解读，可以参阅我写的csdn博客文章 https://blog.csdn.net/nihate/article/details/113850913
when i use cv2.dnn.readNet() to read the onnx model, i have a question
(expected: 'shapes.depth() == CV_32S') where 'shapes.depth()' is 5(CV_32FC1)
must be equal to 'CV_32S' is 4(CV_32SC1) in void_cdecl cv::dnn::dnn4_v20200609::ONNXImporter::populateNet(class cv::dnn::dnn4_v20200609::Net), file D:\opencv_test\newbuild\opencv-master\modules\dnn\src\onnx\innx_importer.cpp, line 1442
[ INFO:0] global C:\jenkins\workspace\OpenCV\OpenVINO\2021.2\build\windows\opencv\modules\dnn\src\onnx\onnx_importer.cpp (395) cv::dnn::dnn4_v20201117::ONNXImporter::populateNet DNN/ONNX: loading ONNX v6 model produced by 'pytorch':1.7. Number of nodes = 335, inputs = 1, outputs = 6
line 103, in get_bboxes_single
max_scores = cls_score.max(axis=1)
numpy.AxisError: axis 1 is out of bounds for array of dimension 1
Hi,
Can you release your training model config file for reference?
Because I use netron check my model is only one output.
But your excellent model have six output.
Have you done any other processing on the model?
Following text is my training config file.
# nanodet-EfficientNet-Lite1_416
save_dir: workspace/RepVGG-A0-416
model:
arch:
name: OneStageDetector
backbone:
name: RepVGG
arch: A0
out_stages: [2,3,4]
activation: ReLU
last_channel: 512
deploy: False
fpn:
name: PAN
in_channels: [96, 192, 512]
out_channels: 128
start_level: 0
num_outs: 3
head:
name: NanoDetHead
num_classes: 1
conv_type: Conv
input_channel: 128
feat_channels: 128
stacked_convs: 2
activation: ReLU
share_cls_reg: True
octave_base_scale: 8
scales_per_octave: 1
strides: [8, 16, 32]
reg_max: 10
norm_cfg:
type: BN
loss:
loss_qfl:
name: QualityFocalLoss
use_sigmoid: True
beta: 2.0
loss_weight: 1.0
loss_dfl:
name: DistributionFocalLoss
loss_weight: 0.25
loss_bbox:
name: GIoULoss
loss_weight: 2.0
data:
train:
name: CocoDataset
img_path: /nfs/Workspace/dataset/mscoco/images/train2017
ann_path: /nfs/Workspace/dataset/mscoco/annotations/annotations/instances_train2017_person.json
input_size: [416,416] #[w,h]
keep_ratio: True
pipeline:
perspective: 0.0
scale: [0.5, 1.5]
stretch: [[1, 1], [1, 1]]
rotation: 0
shear: 0
translate: 0.2
flip: 0.5
brightness: 0.2
contrast: [0.6, 1.4]
saturation: [0.5, 1.2]
normalize: [[103.53, 116.28, 123.675], [57.375, 57.12, 58.395]]
val:
name: CocoDataset
img_path: /nfs/Workspace/dataset/mscoco/images/val2017
ann_path: /nfs/Workspace/dataset/mscoco/annotations/annotations/instances_val2017_person.json
input_size: [416,416] #[w,h]
keep_ratio: True
pipeline:
normalize: [[103.53, 116.28, 123.675], [57.375, 57.12, 58.395]]
device:
gpu_ids: [0]
workers_per_gpu: 1
batchsize_per_gpu: 100
schedule:
# resume:
# load_model: YOUR_MODEL_PATH
optimizer:
name: SGD
lr: 0.07
momentum: 0.9
weight_decay: 0.0001
warmup:
name: linear
steps: 500
ratio: 0.01
total_epochs: 170
lr_schedule:
name: MultiStepLR
milestones: [130,150,160,165]
gamma: 0.1
val_intervals: 5
evaluator:
name: CocoDetectionEvaluator
save_key: mAP
log:
interval: 10
class_names: ['person']
And this is my model architecture output
Here is the model architecture output you provided
Please help! Thanks a lot!
Hello!
Has any plan to use onnxruntime c++ to detect?
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
Django
The Web framework for perfectionists with deadlines.
Laravel
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft
Open source projects and samples from Microsoft.
Google
Google ❤️ Open Source for everyone.
Alibaba
Alibaba Open Source for everyone
D3
Data-Driven Documents codes.
Tencent
China tencent open source team.