Skip to content

README_EN.md

Latest commit

 

History

History
40 lines (30 loc) · 2.33 KB

README_EN.md

File metadata and controls

40 lines (30 loc) · 2.33 KB
Raw

sophon-stream lprnet element

English | 简体中文

The sophon-stream lprnet element is a plugin within the sophon-stream framework, serving as a simple, fast, and powerful license plate recognition model. This project has provided a plugin example for yolo recognition and lprnet license plate detection. For details, please refer to license_plate_recognition.

Features

  • Support for multiple video streams
  • Support for multi-threading

2. Configuration Parameters

The sophon-stream lprnet plugin is divided into three parts: preprocessing, inference, and post-processing, each with configurable parameters that can be adjusted according to specific requirements. Taking inference as an example, here are some commonly used parameters:

{
  "configure": {
    "model_path": "../models/BM1684/lprnet_fp32_1b.bmodel",
    "stage": ["infer"]
  },
  "shared_object": "../../../build/lib/liblprnet.so",
  "name": "lprnet",
  "side": "sophgo",
  "thread_number": 4
}
Parameter Name Type Default Value Description
model_path String "../models/BM1684/lprnet_fp32_1b.bmodel" Path to the lprnet model
stage List ["pre"] Flags for the preprocessing, inference, and post-processing stages
shared_object String "../../../build/lib/liblprnet.so" Path to the liblprnet dynamic library
name String "lprnet" Element name
side String "sophgo" Device type
thread_number Integer 1 Number of threads to launch

Note:

  1. For the stage parameter, it needs to be set to one or a combination of "pre", "infer", "post", and should be connected in order of preprocessing, inference, and post-processing stages. Allocating these three stages to three elements aims to make full use of resources, improving detection efficiency.