FastDeploy/examples/vision/detection/yolov7/python

English | 简体中文

YOLOv7 Python Deployment Demo

Two steps before deployment:

• 1. The hardware and software environment meets the requirements. Please refer to FastDeploy Environment Requirements
• 2. Install FastDeploy Python whl package. Please refer to FastDeploy Python Installation

This doc provides a quick infer.py demo of YOLOv7 deployment on CPU/GPU, and accelerated GPU deployment by TensorRT. Run the following command:

# Download sample deployment code
git clone https://github.com/PaddlePaddle/FastDeploy.git
cd examples/vision/detection/yolov7/python/

# Download yolov7 model files and test images
wget https://bj.bcebos.com/paddlehub/fastdeploy/yolov7.onnx
wget https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/000000014439.jpg

# CPU Inference
python infer.py --model yolov7.onnx --image 000000014439.jpg --device cpu
# GPU
python infer.py --model yolov7.onnx --image 000000014439.jpg --device gpu
# GPU上使用TensorRT推理
python infer.py --model yolov7.onnx --image 000000014439.jpg --device gpu --use_trt True

The visualisation of the results is as follows.

YOLOv7 Python Interface

fastdeploy.vision.detection.YOLOv7(model_file, params_file=None, runtime_option=None, model_format=ModelFormat.ONNX)

YOLOv7 model loading and initialisation, with model_file being the exported ONNX model format.

Parameters

• model_file(str): Model file path
• params_file(str): Parameter file path. If the model format is ONNX, the parameter can be filled with an empty string.
• runtime_option(RuntimeOption): Back-end inference configuration. The default is None, i.e. the default is applied
• model_format(ModelFormat): Model format. The default is ONNX format

Predict Function

YOLOv7.predict(image_data, conf_threshold=0.25, nms_iou_threshold=0.5)

Model prediction interface with direct output of detection results from the image input.

Parameters

• image_data(np.ndarray): Input image. Images need to be in HWC or BGR format
• conf_threshold(float): Filter threshold for detection box confidence
• nms_iou_threshold(float): iou thresholds during NMS processing

Return

Return tofastdeploy.vision.DetectionResultStruct. For more details, please refer to Vision Model Results

Class Member Variables

Pre-processing parameters

Users can modify the following pre-processing parameters for their needs. This will affect the final reasoning and deployment results

• size(list[int]): This parameter modifies the 'resize' during preprocessing and contains two integer elements representing [width, height]. The default value is [640, 640].
• padding_value(list[float]): This parameter modifies the value of the padding when resizing the image. It contains three floating-point elements, representing the values of the three channels. The default value is [114, 114, 114].
• is_no_pad(bool): This parameter determines whether the image is resized by padding, is_no_pad=ture means no padding is used. The default value is is_no_pad=false.
• is_mini_pad(bool): This parameter allows the width and height of the image after resize to be the closest value to the size member variable, which the pixel size of the padding can be divided by the stride member variable. The default value is is_mini_pad=false.
• stride(int): Used withstris_mini_pad member value. The default value isstride=32

Related files

• YOLOv7 Model Introduction
• YOLOv7 C++ Deployment
• Vision Model Results
• how to change inference backend