Golang : Detect number of faces or vehicles in a photo



Tags : golang openCV detect-objects-number detect-faces detect-vehicles

This is a short tutorial on how to use OpenCV to detect the number of faces in an image. Knowing the number of faces in a given photo is crucial in developing machine learning capabilities that involve computer vision.

For examples,

1) A robot needs to know how many people to serve glasses of water.

2) A driverless car needs to know how many incoming cars and number of passengers in the car. IF the drive seat position has a face, chances are the incoming car is in motion.

In this example, we will learn how to detect the number of human faces with a command line Golang program and report the number. If you're looking to detect cars, use the HAAR cascade file for vehicle detection or download the XML files at https://github.com/abhi-kumar/CAR-DETECTION/blob/master/checkcas.xml or http://cogcomp.cs.illinois.edu/Data/Car/

If you need to detect more specific data such as eyes, ears, etc. Use the cascade files at https://github.com/opencv/opencv/tree/master/data/haarcascades.

Here you go!


 package main

 import (
 "fmt"
 "github.com/lazywei/go-opencv/opencv"
 "image"
 gif "image/gif"
 _ "image/jpeg"
 _ "image/png"
 "net/http"
 "os"
 )

 // global variables
 var (
 cascade = new(opencv.HaarCascade)
 )

 func errCheck(err error) {

 if err != nil {
 panic(err)
 }
 }

 func main() {

 if len(os.Args) != 2 {
 fmt.Printf("Usage : %s <imagefilename>\n", os.Args[0])
 os.Exit(0)
 }

 imageFileName := os.Args[1]

 fmt.Println("Analyzing [" + imageFileName + "]......")

 // we will use Go's method of load images
 // instead of openCV.LoadImage
 // because we want to detect if the user supplies animated GIF or not
 imageFile, err := os.Open(imageFileName)

 errCheck(err)

 defer imageFile.Close()

 img, _, err := image.Decode(imageFile)
 errCheck(err)

 buffer := make([]byte, 512)
 imageFile.Seek(0, 0) // reset reader
 _, err = imageFile.Read(buffer)
 errCheck(err)

 filetype := http.DetectContentType(buffer)
 // check if image is GIF and if yes, check to see if it is animated GIF by
 // counting the LoopCount number
 fmt.Println("Analyzing image type : ", filetype)

 if filetype == "image/gif" {
 imageFile.Seek(0, 0)
 // warn if image is animated GIF
 gif, err := gif.DecodeAll(imageFile)
 errCheck(err)
 if gif.LoopCount != 0 {
 fmt.Println("Animated gif detected. Will only scan for faces in the 1st frame.")
 }
 }

 // for vehicles XML files
 // see https://github.com/abhi-kumar/CAR-DETECTION/blob/master/checkcas.xml
 // or http://cogcomp.cs.illinois.edu/Data/Car/
 cascade = opencv.LoadHaarClassifierCascade("./haarcascade_frontalface_alt.xml")
 defer cascade.Release()

 // convert Go's image.Image type to OpenCV's IplImage(Intel Image Processing Library)
 openCVimg := opencv.FromImage(img)
 defer openCVimg.Release()


 if openCVimg != nil {
 faces := cascade.DetectObjects(openCVimg)
 fmt.Printf("Detected [%v] faces in image.\n", len(faces)) // <------- here !

 } else {
 panic("OpenCV FromImage error")
 }

 }

Sample output:

./detectfaces input.gif

Analyzing [input.gif]......

Analyzing image type : image/gif

Detected 1 faces in image.


./detectfaces img.gif

Analyzing [img.gif]......

Analyzing image type : image/gif

Animated gif detected. Will only scan for faces in the 1st frame.

Detected [0] faces in image.

References:

https://github.com/opencv/opencv/tree/master/data/haarcascades

http://docs.opencv.org/3.0-last-rst/doc/tutorials/objdetect/cascadeclassifier/cascadeclassifier.html#cascade-classifier

http://stackoverflow.com/questions/25775650/golang-async-face-detection

Affiliate :

Hire Golang Developers


  See also : Golang : Surveillance with web camera and OpenCV



Tags : golang openCV detect-objects-number detect-faces detect-vehicles

By Adam Ng

IF you gain some knowledge or the information here solved your programming problem. Please consider donating to the less fortunate or some charities that you like. Apart from donation, planting trees, volunteering or reducing your carbon footprint will be great too.


Advertisement