harperjuanl / helmet_yolov5_torchserve

Helmet Detection Deployment on Single Node

Prerequisites

• Python3 and pip3

# check the python version
$ python3 --version
# if not, install as following
$ sudo apt update
$ sudo apt -y upgrade
$ sudo apt install python3.8
$ sudo apt install python3-pip
$ python3 --version

• Docker: Docker Engine comes bundled with Docker Desktop for Linux, this is the easiest and quickest way to get started. If not, you could follow the detailed installation guide

Installation with Docker

$ git clone https://github.com/harperjuanl/helmet_yolov5_torchserve.git
$ cd helmet_yolov5_torchserve
# if you can not clone, you can also download from this link https://github.com/harperjuanl/helmet_yolov5_torchserve/archive/refs/heads/main.zip, and extract the zip file

$ docker build . -t helmet_yolov5_torchserve:v1
$ docker run -p 8080:8080 -p 8081:8081 helmet_yolov5_torchserve:v1

# Register model
$ curl -X POST  "http://localhost:8081/models?url=./model_store/helmet_detection.mar&model_name=helmet_detection"
$ curl -X PUT "http://0.0.0.0:8081/models/helmet_detection?min_worker=3"

Installation without Docker

$ sudo apt-get install openjdk-11-jdk
$ git clone https://github.com/harperjuanl/helmet_yolov5_torchserve.git
$ cd helmet_yolov5_torchserve/resource/helmet_yolov5

$ pip3 install -r requirements.txt -i https://mirrors.aliyun.com/pypi/simple/
$ python3-pip install torchserve
$ cd .. 
$ torchserve --start --ncs --model-store model_store --models helmet_detection.mar

# Register model
$ curl -X POST  "http://localhost:8081/models?url=./model_store/helmet_detection.mar&model_name=helmet_detection"
$ curl -X PUT "http://0.0.0.0:8081/models/helmet_detection?min_worker=3"

Detection & Result

# Detect in the ternimal
$ curl -T test_1.jpg 'http://localhost:8080/predictions/helmet_detection' 
Output: 
[
  {
    "x1": 0.16830310225486755,
    "y1": 0.36698096990585327,
    "x2": 0.3356267809867859,
    "y2": 0.5662754774093628,
    "confidence": 0.9418923854827881,
    "class": "person"
  },
  {
    "x1": -0.0003846943436656147,
    "y1": 0.2697369456291199,
    "x2": 0.11975767463445663,
    "y2": 0.5021408796310425,
    "confidence": 0.9287041425704956,
    "class": "person"
  },
  {
    "x1": 0.31550225615501404,
    "y1": 0.27130556106567383,
    "x2": 0.4195330739021301,
    "y2": 0.4244980812072754,
    "confidence": 0.9224411249160767,
    "class": "person"
  },
  {
    "x1": 0.8000054359436035,
    "y1": 0.36035841703414917,
    "x2": 0.8742903470993042,
    "y2": 0.4628569483757019,
    "confidence": 0.9012498259544373,
    "class": "person"
  },
  {
    "x1": 0.44192060828208923,
    "y1": 0.3977605700492859,
    "x2": 0.5190550088882446,
    "y2": 0.4892307221889496,
    "confidence": 0.8915991187095642,
    "class": "person"
  },
  {
    "x1": 0.9677120447158813,
    "y1": 0.4071219563484192,
    "x2": 0.9998529553413391,
    "y2": 0.5111279487609863,
    "confidence": 0.875587522983551,
    "class": "person"
  },
  {
    "x1": 0.5246236324310303,
    "y1": 0.39841872453689575,
    "x2": 0.5718141794204712,
    "y2": 0.4656790792942047,
    "confidence": 0.8437989950180054,
    "class": "hat"
  },
  {
    "x1": 0.6443458795547485,
    "y1": 0.2609959542751312,
    "x2": 0.7564457654953003,
    "y2": 0.443324476480484,
    "confidence": 0.84254390001297,
    "class": "person"
  },
  {
    "x1": 0.6181862950325012,
    "y1": 0.3655022084712982,
    "x2": 0.6777603030204773,
    "y2": 0.452880322933197,
    "confidence": 0.7514644861221313,
    "class": "person"
  }
]%


# You can also choose to detect with the jupyter notebook
$ python3 -m pip install virtualenv
$ python3 -m virtualenv helmet-env
$ source yolov5-env/bin/activate

$ pip3 install -r requirements.txt
$ python3 -m pip install jupyter      # only need to install once if you don't have jupyter 
$ jupyter notebook 
# open the /helmet_yolov5_torchserve/resource/pytorch-yolov5-helmet-detection-inference.ipynb and run the code cell


# Stop the model serving
$ docker container ls
$ docker stop [the helmet container CONTAINER ID]