Severe Memory Leak During Conversion between Image and gocv.Mat

[Sclock]

Description

In this issue, it is reported that there is a severe memory leak during the conversion between image.Image and gocv.Mat types, specifically when these conversions are done within a goroutine pool. Despite the fact that all the images are reported as Empty, a significant amount

Steps to Reproduce

package main  
  
import (  
	"fmt"  
	"image"  
	"time"  
  
	"github.com/panjf2000/ants"  
	"gocv.io/x/gocv"  
)  
  
// MatToMat converts a gocv.Mat to another gocv.Mat through an image representation  
func MatToMat(mat gocv.Mat) gocv.Mat {  
	// Convert gocv.Mat to image.Image  
	img, _ := mat.ToImage()  
  
	// Convert image.Image back to gocv.Mat (RGB format)  
	reImg, _ := gocv.ImageToMatRGB(img)  
  
	// Ensure the new gocv.Mat is closed to prevent memory leaks  
	defer reImg.Close()  
  
	return reImg  
}  
  
// ImgToImg converts an image.Image to another image.Image through a gocv.Mat representation  
func ImgToImg(img image.Image) image.Image {  
	// Convert image.Image to gocv.Mat (RGB format)  
	mat, _ := gocv.ImageToMatRGB(img)  
  
	// Convert gocv.Mat back to image.Image  
	reImg, _ := mat.ToImage()  
  
	return reImg  
}  
  
// imgInfo holds an image's gocv.Mat representation and its name  
type imgInfo struct {  
	img  gocv.Mat  
	name string  
}  
  
// sendGocv is a channel used to send imgInfo instances  
var sendGocv = make(chan imgInfo)  
  
// imgdict is a map that stores imgInfo instances  
var imgdict = make(map[string]gocv.Mat)  
  
// poolWait waits for all images to be processed and stored in imgdict  
func poolWait(count int) {  
	for {  
		select {  
		case imginfo := <-sendGocv:  
			imgdict[imginfo.name] = imginfo.img  
			count--  
			// fmt.Println("Remaining", count, "Processed", len(imgdict))  
			if count == 0 {  
				return  
			}  
		}  
	}  
}  
  
// createAndRunPool creates a goroutine pool to run a specified task multiple times  
func createAndRunPool(count, poolSize int, task func(text string)) {  
	p, err := ants.NewPool(poolSize)  
	if err != nil {  
		panic(err)  
	}  
	defer p.Release()  
  
	go func() {  
		for i := 0; i < count; i++ {  
			err = p.Submit(func() {  
				msg := fmt.Sprintf("%d", i)  
				task(msg)  
			})  
			if err != nil {  
				panic(err)  
			}  
		}  
	}()  
	poolWait(count)  
}  
  
func main() {  
	file := "./any_png.png"  
	runCount := 100  
  
	mat := gocv.IMRead(file, gocv.IMReadUnchanged)  
  
	fmt.Println("first")  
	createAndRunPool(runCount, 20, func(text string) {  
		newMat := MatToMat(mat)  
		sendGocv <- imgInfo{newMat, text}  
	})  
	fmt.Println("end")  
  
	time.Sleep(5 * time.Second)  
  
	fmt.Println("Second round")  
	createAndRunPool(runCount, 20, func(text string) {  
		newMat := MatToMat(mat)  
		sendGocv <- imgInfo{newMat, text}  
	})  
	fmt.Println("End")  
  
	time.Sleep(5 * time.Second)  
  
	for _, ig := range imgdict {  
		if ig.Empty() {  
			fmt.Println("Empty!")  
		}  
	}  
}

Your Environment

• Operating System and version: Linux 5.4.0-126-generic Feature/mat pointer #142-Ubuntu SMP Fri Aug 26 12:12:57 UTC 2022 x86_64 x86_64 x86_64 GNU/Linux
• OpenCV version used: 4.8.1
• How did you install OpenCV? make build
• GoCV version used: v0.35.0
• Go version: 1.22.1
• Did you run the env.sh or env.cmd script before trying to go run or go build?

If my usage method is incorrect, please tell me the correct usage. Thank you

[deadprogram]

Hello @Sclock please see https://github.com/hybridgroup/gocv?tab=readme-ov-file#profiling for explanation.