apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-22 16:07:51 +08:00
Code Issues Actions 15 Packages Projects Releases Wiki Activity

[Model] Add YOLOV8 For RKNPU2 (#1153)

Browse Source 
* 更新ppdet

* 更新ppdet

* 更新ppdet

* 更新ppdet

* 更新ppdet

* 新增ppdet_decode

* 更新多batch支持

* 更新多batch支持

* 更新多batch支持

* 更新注释内容

* 尝试解决pybind问题

* 尝试解决pybind的问题

* 尝试解决pybind的问题

* 重构代码

* 重构代码

* 重构代码

* 按照要求修改

* 更新Picodet文档

* 更新Picodet文档，更新yolov8文档

* 修改picodet 以及 yolov8 example

* 更新Picodet模型转换脚本

* 更新example代码

* 更新yolov8量化代码

* 修复部分bug
加入pybind

* 修复pybind

* 修复pybind错误的问题

* 更新说明文档

* 更新说明文档
This commit is contained in:
Zheng-Bicheng
2023-01-16 22:33:02 +08:00
committed by GitHub
parent 66240a6f66
commit f441ffe56b
12 changed files with 290 additions and 189 deletions
Download Patch File Download Diff File Expand all files Collapse all files
examples/vision/detection/paddledetection/rknpu2/README_CN.md
+12 -102
View File
@@ -1,121 +1,31 @@
[English](README.md) | 简体中文
# PaddleDetection RKNPU2部署示例
## 支持模型列表
目前FastDeploy支持如下模型的部署
- [PicoDet系列模型](https://github.com/PaddlePaddle/PaddleDetection/tree/release/2.4/configs/picodet)
目前FastDeploy使用RKNPU2支持如下PaddleDetection模型的部署:
- Picodet
- PPYOLOE
- YOLOV8
## 准备PaddleDetection部署模型以及转换模型
RKNPU部署模型前需要将Paddle模型转换成RKNN模型，具体步骤如下:
* Paddle动态图模型转换为ONNX模型，请参考[PaddleDetection导出模型](https://github.com/PaddlePaddle/PaddleDetection/blob/release/2.4/deploy/EXPORT_MODEL.md)
,注意在转换时请设置**export.nms=True**.
,注意在转换时请设置**export.nms=True**.
* ONNX模型转换RKNN模型的过程，请参考[转换文档](../../../../../docs/cn/faq/rknpu2/export.md)进行转换。
## 模型转换example
以下步骤均在Ubuntu电脑上完成，请参考配置文档完成转换模型环境配置。下面以Picodet-s为例子,教大家如何转换PaddleDetection模型到RKNN模型。
### 导出ONNX模型
```bash
# 下载Paddle静态图模型并解压
wget https://paddledet.bj.bcebos.com/deploy/Inference/picodet_s_416_coco_lcnet.tar
tar xvf picodet_s_416_coco_lcnet.tar
- [Picodet RKNPU2模型转换文档](./picodet.md)
- [YOLOv8 RKNPU2模型转换文档](./yolov8.md)
# 静态图转ONNX模型，注意，这里的save_file请和压缩包名对齐
paddle2onnx --model_dir picodet_s_416_coco_lcnet \
--model_filename model.pdmodel \
--params_filename model.pdiparams \
--save_file picodet_s_416_coco_lcnet/picodet_s_416_coco_lcnet.onnx \
--enable_dev_version True
# 固定shape
python -m paddle2onnx.optimize --input_model picodet_s_416_coco_lcnet/picodet_s_416_coco_lcnet.onnx \
--output_model picodet_s_416_coco_lcnet/picodet_s_416_coco_lcnet.onnx \
--input_shape_dict "{'image':[1,3,416,416]}"
```
### 编写模型导出配置文件
以转化RK3568的RKNN模型为例子，我们需要编辑tools/rknpu2/config/RK3568/picodet_s_416_coco_lcnet.yaml，来转换ONNX模型到RKNN模型。
**修改normalize参数**
如果你需要在NPU上执行normalize操作，请根据你的模型配置normalize参数，例如:
```yaml
model_path: ./picodet_s_416_coco_lcnet/picodet_s_416_coco_lcnet.onnx
output_folder: ./picodet_s_416_coco_lcnet
target_platform: RK3568
normalize:
mean: [[0.485,0.456,0.406]]
std: [[0.229,0.224,0.225]]
outputs: ['tmp_17','p2o.Concat.9']
```
**修改outputs参数**
由于Paddle2ONNX版本的不同，转换模型的输出节点名称也有所不同，请使用[Netron](https://netron.app)，并找到以下蓝色方框标记的NonMaxSuppression节点，红色方框的节点名称即为目标名称。
例如，使用Netron可视化后，得到以下图片:
![](https://user-images.githubusercontent.com/58363586/202728663-4af0b843-d012-4aeb-8a66-626b7b87ca69.png)
找到蓝色方框标记的NonMaxSuppression节点，可以看到红色方框标记的两个节点名称为tmp_17和p2o.Concat.9,因此需要修改outputs参数，修改后如下:
```yaml
model_path: ./picodet_s_416_coco_lcnet/picodet_s_416_coco_lcnet.onnx
output_folder: ./picodet_s_416_coco_lcnet
target_platform: RK3568
normalize: None
outputs: ['tmp_17','p2o.Concat.9']
```
### 转换模型
```bash
# ONNX模型转RKNN模型
# 转换模型,模型将生成在picodet_s_320_coco_lcnet_non_postprocess目录下
python tools/rknpu2/export.py --config_path tools/rknpu2/config/picodet_s_416_coco_lcnet.yaml \
--target_platform rk3588
```
### 修改模型运行时的配置文件
配置文件中，我们只需要修改**Preprocess**下的**Normalize**和**Permute**.
**删除Permute**
RKNPU只支持NHWC的输入格式，因此需要删除Permute操作.删除后，配置文件Precess部分后如下:
```yaml
Preprocess:
- interp: 2
keep_ratio: false
target_size:
- 416
- 416
type: Resize
- is_scale: true
mean:
- 0.485
- 0.456
- 0.406
std:
- 0.229
- 0.224
- 0.225
type: NormalizeImage
```
**根据模型转换文件决定是否删除Normalize**
RKNPU支持使用NPU进行Normalize操作，如果你在导出模型时配置了Normalize参数，请删除**Normalize**.删除后配置文件Precess部分如下:
```yaml
Preprocess:
- interp: 2
keep_ratio: false
target_size:
- 416
- 416
type: Resize
```
## 其他链接
- [Cpp部署](./cpp)
- [Python部署](./python)
- [视觉模型预测结果](../../../../../docs/api/vision_results/)
examples/vision/detection/paddledetection/rknpu2/cpp/CMakeLists.txt
+10 -31
View File
@@ -1,37 +1,16 @@
CMAKE_MINIMUM_REQUIRED(VERSION 3.10)
project(rknpu2_test)
set(CMAKE_CXX_STANDARD 14)
PROJECT(infer_demo C CXX)
CMAKE_MINIMUM_REQUIRED (VERSION 3.10)
# 指定下载解压后的fastdeploy库路径
set(FASTDEPLOY_INSTALL_DIR "thirdpartys/fastdeploy-0.0.3")
option(FASTDEPLOY_INSTALL_DIR "Path of downloaded fastdeploy sdk.")
include(${FASTDEPLOY_INSTALL_DIR}/FastDeployConfig.cmake)
include_directories(${FastDeploy_INCLUDE_DIRS})
include(${FASTDEPLOY_INSTALL_DIR}/FastDeploy.cmake)
add_executable(infer_picodet infer_picodet.cc)
target_link_libraries(infer_picodet ${FastDeploy_LIBS})
# 添加FastDeploy依赖头文件
include_directories(${FASTDEPLOY_INCS})
add_executable(infer_picodet_demo ${PROJECT_SOURCE_DIR}/infer_picodet_demo.cc)
target_link_libraries(infer_picodet_demo ${FASTDEPLOY_LIBS})
set(CMAKE_INSTALL_PREFIX ${CMAKE_SOURCE_DIR}/build/install)
install(TARGETS infer_picodet DESTINATION ./)
install(DIRECTORY model DESTINATION ./)
install(DIRECTORY images DESTINATION ./)
file(GLOB FASTDEPLOY_LIBS ${FASTDEPLOY_INSTALL_DIR}/lib/*)
message("${FASTDEPLOY_LIBS}")
install(PROGRAMS ${FASTDEPLOY_LIBS} DESTINATION lib)
file(GLOB ONNXRUNTIME_LIBS ${FASTDEPLOY_INSTALL_DIR}/third_libs/install/onnxruntime/lib/*)
install(PROGRAMS ${ONNXRUNTIME_LIBS} DESTINATION lib)
install(DIRECTORY ${FASTDEPLOY_INSTALL_DIR}/third_libs/install/opencv/lib DESTINATION ./)
file(GLOB PADDLETOONNX_LIBS ${FASTDEPLOY_INSTALL_DIR}/third_libs/install/paddle2onnx/lib/*)
install(PROGRAMS ${PADDLETOONNX_LIBS} DESTINATION lib)
file(GLOB RKNPU2_LIBS ${FASTDEPLOY_INSTALL_DIR}/third_libs/install/rknpu2_runtime/${RKNN2_TARGET_SOC}/lib/*)
install(PROGRAMS ${RKNPU2_LIBS} DESTINATION lib)
add_executable(infer_yolov8_demo ${PROJECT_SOURCE_DIR}/infer_yolov8_demo.cc)
target_link_libraries(infer_yolov8_demo ${FASTDEPLOY_LIBS})
examples/vision/detection/paddledetection/rknpu2/cpp/README_CN.md
+18 -41
View File
@@ -1,4 +1,5 @@
[English](README.md) | 简体中文
# PaddleDetection C++部署示例
本目录下提供`infer_picodet.cc`快速完成PPDetection模型在Rockchip板子上上通过二代NPU加速部署的示例。
@@ -10,50 +11,24 @@
以上步骤请参考[RK2代NPU部署库编译](../../../../../../docs/cn/build_and_install/rknpu2.md)实现
## 生成基本目录文件
该例程由以下几个部分组成
```text
.
├── CMakeLists.txt
├── build # 编译文件夹
├── image # 存放图片的文件夹
├── infer_picodet.cc
├── model # 存放模型文件的文件夹
└── thirdpartys # 存放sdk的文件夹
```
首先需要先生成目录结构
```bash
以picodet为例进行推理部署
mkdir build
mkdir images
mkdir model
mkdir thirdpartys
```
## 编译
### 编译并拷贝SDK到thirdpartys文件夹
请参考[RK2代NPU部署库编译](../../../../../../docs/cn/build_and_install/rknpu2.md)仓库编译SDK，编译完成后，将在build目录下生成fastdeploy-0.0.3目录，请移动它至thirdpartys目录下.
### 拷贝模型文件，以及配置文件至model文件夹
在Paddle动态图模型 -> Paddle静态图模型 -> ONNX模型的过程中，将生成ONNX文件以及对应的yaml配置文件，请将配置文件存放到model文件夹内。
转换为RKNN后的模型文件也需要拷贝至model。
### 准备测试图片至image文件夹
```bash
wget https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/000000014439.jpg
cp 000000014439.jpg ./images
```
### 编译example
```bash
cd build
cmake ..
make -j8
make install
# 下载预编译库，详情见文档导航处
wget https://bj.bcebos.com/fastdeploy/release/cpp/fastdeploy-linux-x64-x.x.x.tgz
tar xvf fastdeploy-linux-x64-x.x.x.tgz
cmake .. -DFASTDEPLOY_INSTALL_DIR=${PWD}/fastdeploy-linux-x64-x.x.x
make -j
# 下载PPYOLOE模型文件和测试图片
wget https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/000000014439.jpg
# CPU推理
./infer_picodet_demo ./picodet_s_416_coco_lcnet 000000014439.jpg 0
# RKNPU2推理
./infer_picodet_demo ./picodet_s_416_coco_lcnet 000000014439.jpg 1
```
## 运行例程
@@ -63,7 +38,9 @@ cd ./build/install
./infer_picodet model/picodet_s_416_coco_lcnet images/000000014439.jpg
```
## 文档导航
- [模型介绍](../../)
- [Python部署](../python)
- [视觉模型预测结果](../../../../../../docs/api/vision_results/)
- [RKNPU2 预编译库](../../../../../../docs/cn/faq/rknpu2/rknpu2.md)
examples/vision/detection/paddledetection/rknpu2/cpp/infer_picodet.cc
+15 -12
View File
@@ -13,8 +13,8 @@
// limitations under the License.
#include <iostream>
#include <string>
#include "fastdeploy/vision.h"
#include <sys/time.h>
void ONNXInfer(const std::string& model_dir, const std::string& image_file) {
std::string model_file = model_dir + "/picodet_s_416_coco_lcnet.onnx";
@@ -25,7 +25,7 @@ void ONNXInfer(const std::string& model_dir, const std::string& image_file) {
auto format = fastdeploy::ModelFormat::ONNX;
auto model = fastdeploy::vision::detection::PicoDet(
model_file, params_file, config_file,option,format);
model_file, params_file, config_file, option, format);
fastdeploy::TimeCounter tc;
tc.Start();
@@ -35,14 +35,12 @@ void ONNXInfer(const std::string& model_dir, const std::string& image_file) {
std::cerr << "Failed to predict." << std::endl;
return;
}
auto vis_im = fastdeploy::vision::VisDetection(im, res,0.5);
auto vis_im = fastdeploy::vision::VisDetection(im, res, 0.5);
tc.End();
tc.PrintInfo("PPDet in ONNX");
cv::imwrite("infer_onnx.jpg", vis_im);
std::cout
<< "Visualized result saved in ./infer_onnx.jpg"
<< std::endl;
std::cout << "Visualized result saved in ./infer_onnx.jpg" << std::endl;
}
void RKNPU2Infer(const std::string& model_dir, const std::string& image_file) {
@@ -56,8 +54,10 @@ void RKNPU2Infer(const std::string& model_dir, const std::string& image_file) {
auto format = fastdeploy::ModelFormat::RKNN;
auto model = fastdeploy::vision::detection::PicoDet(
model_file, params_file, config_file,option,format);
model_file, params_file, config_file, option, format);
model.GetPreprocessor().DisablePermute();
model.GetPreprocessor().DisableNormalize();
model.GetPostprocessor().ApplyDecodeAndNMS();
auto im = cv::imread(image_file);
@@ -73,21 +73,24 @@ void RKNPU2Infer(const std::string& model_dir, const std::string& image_file) {
tc.PrintInfo("PPDet in RKNPU2");
std::cout << res.Str() << std::endl;
auto vis_im = fastdeploy::vision::VisDetection(im, res,0.5);
auto vis_im = fastdeploy::vision::VisDetection(im, res, 0.5);
cv::imwrite("infer_rknpu2.jpg", vis_im);
std::cout << "Visualized result saved in ./infer_rknpu2.jpg" << std::endl;
}
int main(int argc, char* argv[]) {
if (argc < 3) {
if (argc < 4) {
std::cout
<< "Usage: infer_demo path/to/model_dir path/to/image run_option, "
"e.g ./infer_model ./picodet_model_dir ./test.jpeg"
<< std::endl;
return -1;
}
RKNPU2Infer(argv[1], argv[2]);
//ONNXInfer(argv[1], argv[2]);
if (std::atoi(argv[3]) == 0) {
ONNXInfer(argv[1], argv[2]);
} else if (std::atoi(argv[3]) == 1) {
RKNPU2Infer(argv[1], argv[2]);
}
return 0;
}
examples/vision/detection/paddledetection/rknpu2/cpp/infer_yolov8.cc
+95
View File
@@ -0,0 +1,95 @@
// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/vision.h"
void ONNXInfer(const std::string& model_dir, const std::string& image_file) {
std::string model_file = model_dir + "/yolov8_n_500e_coco.onnx";
std::string params_file;
std::string config_file = model_dir + "/infer_cfg.yml";
auto option = fastdeploy::RuntimeOption();
option.UseCpu();
auto format = fastdeploy::ModelFormat::ONNX;
auto model = fastdeploy::vision::detection::PaddleYOLOv8(
model_file, params_file, config_file, option, format);
fastdeploy::TimeCounter tc;
tc.Start();
auto im = cv::imread(image_file);
fastdeploy::vision::DetectionResult res;
if (!model.Predict(im, &res)) {
std::cerr << "Failed to predict." << std::endl;
return;
}
auto vis_im = fastdeploy::vision::VisDetection(im, res, 0.5);
tc.End();
tc.PrintInfo("PPDet in ONNX");
std::cout << res.Str() << std::endl;
cv::imwrite("infer_onnx.jpg", vis_im);
std::cout << "Visualized result saved in ./infer_onnx.jpg" << std::endl;
}
void RKNPU2Infer(const std::string& model_dir, const std::string& image_file) {
auto model_file = model_dir + "/yolov8_n_500e_coco_rk3588_unquantized.rknn";
auto params_file = "";
auto config_file = model_dir + "/infer_cfg.yml";
auto option = fastdeploy::RuntimeOption();
option.UseRKNPU2();
auto format = fastdeploy::ModelFormat::RKNN;
auto model = fastdeploy::vision::detection::PaddleYOLOv8(
model_file, params_file, config_file, option, format);
model.GetPreprocessor().DisablePermute();
model.GetPreprocessor().DisableNormalize();
model.GetPostprocessor().ApplyDecodeAndNMS();
auto im = cv::imread(image_file);
fastdeploy::vision::DetectionResult res;
fastdeploy::TimeCounter tc;
tc.Start();
if (!model.Predict(&im, &res)) {
std::cerr << "Failed to predict." << std::endl;
return;
}
tc.End();
tc.PrintInfo("PPDet in RKNPU2");
std::cout << res.Str() << std::endl;
auto vis_im = fastdeploy::vision::VisDetection(im, res, 0.5);
cv::imwrite("infer_rknpu2.jpg", vis_im);
std::cout << "Visualized result saved in ./infer_rknpu2.jpg" << std::endl;
}
int main(int argc, char* argv[]) {
if (argc < 4) {
std::cout
<< "Usage: infer_demo path/to/model_dir path/to/image run_option, "
"e.g ./infer_model ./picodet_model_dir ./test.jpeg"
<< std::endl;
return -1;
}
if (std::atoi(argv[3]) == 0) {
ONNXInfer(argv[1], argv[2]);
} else if (std::atoi(argv[3]) == 1) {
RKNPU2Infer(argv[1], argv[2]);
}
return 0;
}
examples/vision/detection/paddledetection/rknpu2/picodet.md
+68
View File
@@ -0,0 +1,68 @@
# Picodet RKNPU2模型转换文档
以下步骤均在Ubuntu电脑上完成，请参考配置文档完成转换模型环境配置。下面以Picodet-s为例子,教大家如何转换PaddleDetection模型到RKNN模型。
### 导出ONNX模型
```bash
# 下载Paddle静态图模型并解压
wget https://paddledet.bj.bcebos.com/deploy/Inference/picodet_s_416_coco_lcnet.tar
tar xvf picodet_s_416_coco_lcnet.tar
# 静态图转ONNX模型，注意，这里的save_file请和压缩包名对齐
paddle2onnx --model_dir picodet_s_416_coco_lcnet \
--model_filename model.pdmodel \
--params_filename model.pdiparams \
--save_file picodet_s_416_coco_lcnet/picodet_s_416_coco_lcnet.onnx \
--enable_dev_version True
# 固定shape
python -m paddle2onnx.optimize --input_model picodet_s_416_coco_lcnet/picodet_s_416_coco_lcnet.onnx \
--output_model picodet_s_416_coco_lcnet/picodet_s_416_coco_lcnet.onnx \
--input_shape_dict "{'image':[1,3,416,416]}"
```
### 编写模型导出配置文件
以转化RK3568的RKNN模型为例子，我们需要编辑tools/rknpu2/config/picodet_s_416_coco_lcnet_unquantized.yaml，来转换ONNX模型到RKNN模型。
**修改normalize参数**
如果你需要在NPU上执行normalize操作，请根据你的模型配置normalize参数，例如:
```yaml
mean:
-
- 127.5
- 127.5
- 127.5
std:
-
- 127.5
- 127.5
- 127.5
```
**修改outputs参数**
由于Paddle2ONNX版本的不同，转换模型的输出节点名称也有所不同，请使用[Netron](https://netron.app)对模型进行可视化，并找到以下蓝色方框标记的NonMaxSuppression节点，红色方框的节点名称即为目标名称。
例如，使用Netron可视化后，得到以下图片:
![](https://user-images.githubusercontent.com/58363586/212599781-e1952da7-6eae-4951-8ca7-bab7e6940692.png)
找到蓝色方框标记的NonMaxSuppression节点，可以看到红色方框标记的两个节点名称为p2o.Div.79和p2o.Concat.9,因此需要修改outputs参数，修改后如下:
```yaml
outputs_nodes: [ 'p2o.Div.79','p2o.Concat.9' ]
```
### 转换模型
```bash
# ONNX模型转RKNN模型
# 转换模型,模型将生成在picodet_s_320_coco_lcnet_non_postprocess目录下
python tools/rknpu2/export.py --config_path tools/rknpu2/config/picodet_s_416_coco_lcnet_unquantized.yaml \
--target_platform rk3588
```
examples/vision/detection/paddledetection/rknpu2/yolov8.md
+50
View File
@@ -0,0 +1,50 @@
# YOLOv8 RKNPU2模型转换文档
以下步骤均在Ubuntu电脑上完成，请参考配置文档完成转换模型环境配置。下面以yolov8为例子,教大家如何转换PaddleDetection模型到RKNN模型。
### 导出ONNX模型
```bash
# 下载Paddle静态图模型并解压
# 静态图转ONNX模型，注意，这里的save_file请和压缩包名对齐
paddle2onnx --model_dir yolov8_n_500e_coco \
--model_filename model.pdmodel \
--params_filename model.pdiparams \
--save_file yolov8_n_500e_coco/yolov8_n_500e_coco.onnx \
--enable_dev_version True
# 固定shape
python -m paddle2onnx.optimize --input_model yolov8_n_500e_coco/yolov8_n_500e_coco.onnx \
--output_model yolov8_n_500e_coco/yolov8_n_500e_coco.onnx \
--input_shape_dict "{'image':[1,3,640,640],'scale_factor':[1,2]}"
```
### 编写模型导出配置文件
**修改outputs参数**
由于Paddle2ONNX版本的不同，转换模型的输出节点名称也有所不同，请使用[Netron](https://netron.app)对模型进行可视化，并找到以下蓝色方框标记的NonMaxSuppression节点，红色方框的节点名称即为目标名称。
例如，使用Netron可视化后，得到以下图片:
![](https://user-images.githubusercontent.com/58363586/212599658-8a2c4b79-f59a-40b5-ade7-f77c6fcfdf2a.png)
找到蓝色方框标记的NonMaxSuppression节点，可以看到红色方框标记的两个节点名称为p2o.Div.1和p2o.Concat.9,因此需要修改outputs参数，修改后如下:
```yaml
outputs_nodes: [ 'p2o.Div.1','p2o.Concat.49' ]
```
### 转换模型
```bash
# ONNX模型转RKNN模型
# 转换非全量化模型,模型将生成在yolov8_n目录下
python tools/rknpu2/export.py --config_path tools/rknpu2/config/yolov8_n_unquantized.yaml \
--target_platform rk3588
# 转换全量化模型,模型将生成在yolov8_n目录下
python tools/rknpu2/export.py --config_path tools/rknpu2/config/yolov8_n_quantized.yaml \
--target_platform rk3588
```
fastdeploy/vision/detection/ppdet/model.h
Executable → Regular
+1
View File
@@ -258,6 +258,7 @@ class FASTDEPLOY_DECL PaddleYOLOv8 : public PPDetBase {
Backend::LITE};
valid_gpu_backends = {Backend::ORT, Backend::PDINFER, Backend::TRT};
valid_kunlunxin_backends = {Backend::LITE};
valid_rknpu_backends = {Backend::RKNPU2};
valid_ascend_backends = {Backend::LITE};
initialized = Initialize();
}
fastdeploy/vision/detection/ppdet/preprocessor.cc
-2
View File
@@ -90,11 +90,9 @@ bool PaddleDetPreprocessor::BuildPreprocessPipelineFromConfig() {
// Do nothing, do permute as the last operation
has_permute = true;
continue;
// processors_.push_back(std::make_shared<HWC2CHW>());
} else if (op_name == "Pad") {
auto size = op["size"].as<std::vector<int>>();
auto value = op["fill_value"].as<std::vector<float>>();
processors_.push_back(std::make_shared<Cast>("float"));
processors_.push_back(
std::make_shared<PadToSize>(size[1], size[0], value));
} else if (op_name == "PadStride") {
tools/rknpu2/config/picodet_s_416_coco_lcnet.yaml → tools/rknpu2/config/picodet_s_416_coco_lcnet_unquantized.yaml
+3 -1
View File
@@ -10,6 +10,8 @@ std:
- 127.5
model_path: ./picodet_s_416_coco_lcnet/picodet_s_416_coco_lcnet.onnx
outputs_nodes:
- 'p2o.Div.79'
- 'p2o.Concat.9'
do_quantization: False
dataset:
output_folder: "./Portrait_PP_HumanSegV2_Lite_256x144_infer"
output_folder: "./picodet_s_416_coco_lcnet"
tools/rknpu2/config/yolov8_n_quantized.yaml
+9
View File
@@ -0,0 +1,9 @@
mean:
std:
model_path: ./yolov8_n_500e_coco/yolov8_n_500e_coco.onnx
outputs_nodes:
- 'p2o.Mul.119'
- 'p2o.Concat.49'
do_quantization: True
dataset: "./yolov8_n_500e_coco/dataset.txt"
output_folder: "./yolov8_n_500e_coco"
tools/rknpu2/config/yolov8_n_unquantized.yaml
+9
View File
@@ -0,0 +1,9 @@
mean:
std:
model_path: ./yolov8_n_500e_coco/yolov8_n_500e_coco.onnx
outputs_nodes:
- 'p2o.Div.1'
- 'p2o.Concat.49'
do_quantization: False
dataset: "./dataset.txt"
output_folder: "./yolov8_n_500e_coco"