English | 简体中文
This directory provides examples that infer.cc fast finishes the deployment of RetinaFace on CPU/GPU and GPU accelerated by TensorRT.
Before deployment, two steps require confirmation
-
- Software and hardware should meet the requirements. Please refer to FastDeploy Environment Requirements
-
- Download the precompiled deployment library and samples code according to your development environment. Refer to FastDeploy Precompiled Library
Taking the CPU inference on Linux as an example, the compilation test can be completed by executing the following command in this directory. FastDeploy version 0.7.0 or above (x.x.x>=0.7.0) is required to support this model.
mkdir build
cd build
# Download the FastDeploy precompiled library. Users can choose your appropriate version in the `FastDeploy Precompiled Library` mentioned above
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
# Download RetinaFace model files and test images
wget https://bj.bcebos.com/paddlehub/fastdeploy/Pytorch_RetinaFace_mobile0.25-640-640.onnx
wget https://raw.githubusercontent.com/DefTruth/lite.ai.toolkit/main/examples/lite/resources/test_lite_face_detector_3.jpg
# CPU inference
./infer_demo Pytorch_RetinaFace_mobile0.25-640-640.onnx test_lite_face_detector_3.jpg 0
# GPU inference
./infer_demo Pytorch_RetinaFace_mobile0.25-640-640.onnx test_lite_face_detector_3.jpg 1
# TensorRT inference on GPU
./infer_demo Pytorch_RetinaFace_mobile0.25-640-640.onnx test_lite_face_detector_3.jpg 2The visualized result after running is as follows
The above command works for Linux or MacOS. For SDK use-pattern in Windows, refer to:
fastdeploy::vision::facedet::RetinaFace(
const string& model_file,
const string& params_file = "",
const RuntimeOption& runtime_option = RuntimeOption(),
const ModelFormat& model_format = ModelFormat::ONNX)RetinaFace model loading and initialization, among which model_file is the exported ONNX model format
Parameter
- model_file(str): Model file path
- params_file(str): Parameter file path. Only passing an empty string when the model is in ONNX format
- runtime_option(RuntimeOption): Backend inference configuration. None by default, which is the default configuration
- model_format(ModelFormat): Model format. ONNX format by default
RetinaFace::Predict(cv::Mat* im, FaceDetectionResult* result, float conf_threshold = 0.25, float nms_iou_threshold = 0.5)Model prediction interface. Input images and output detection results.
Parameter
- im: Input images in HWC or BGR format
- result: Detection results, including detection box and confidence of each box. Refer to Vision Model Prediction Result for FaceDetectionResult
- conf_threshold: Filtering threshold of detection box confidence
- nms_iou_threshold: iou threshold during NMS processing
Users can modify the following pre-processing parameters to their needs, which affects the final inference and deployment results
- size(vector<int>): This parameter changes the size of the resize used during preprocessing, containing two integer elements for [width, height] with default value [640, 640]
- variance(vector<float>): This parameter is used to change the padding value of images during resize, containing three floating-point elements that represent the value of three channels. Default value [0, 0, 0]
- min_sizes(vector<vector<int>>): Set width and height of anchor in retinaface. Default {{16, 32}, {64, 128}, {256, 512}}, corresponding to step size 8, 16 and 32
- downsample_strides(vector<int>): This parameter is used to change the down-sampling multiple of the feature map that generates anchor, containing three integer elements that represent the default down-sampling multiple for generating anchor. Default value [8, 16, 32]
- landmarks_per_face(int): Specify the number of keypoints in the face detected. Default 5