Auto annotation for plain background images.

auto_annotate.py

import cv2
import numpy as np
from matplotlib import pyplot as plt

def get_bbox(path, plot=False):
    img = cv2.imread(path)
    orig = img.copy()

    ################ 
    # denoise
    ###############
    dn = cv2.fastNlMeansDenoisingColored(img,None,30,7,21)

    ############
    # mask
    ############
    hsv = cv2.cvtColor(dn, cv2.COLOR_BGR2HSV)
    # green color mask
    mask = cv2.inRange(hsv, (26, 25, 25), (70, 255,255))

    ## slice the green
    imask = mask>0
    mask = np.zeros_like(img, np.uint8)
    mask.fill(255)
    mask[imask] = img[imask]

    ###########
    # K means
    ###########
    Z = mask.reshape((-1,3))
    Z = np.float32(Z)
    # define criteria, number of clusters(K) and apply kmeans()
    criteria = (cv2.TERM_CRITERIA_EPS + cv2.TERM_CRITERIA_MAX_ITER, 10, 1.0)
    K = 2
    ret,label,center=cv2.kmeans(Z,K,None,criteria,10,cv2.KMEANS_RANDOM_CENTERS)
    center = np.uint8(center)
    res = center[label.flatten()]
    km = res.reshape((img.shape))

    ################
    # bw from km
    ################
    gray = cv2.cvtColor(km,cv2.COLOR_BGR2GRAY)
    thresh = 127
    bw = cv2.threshold(gray, thresh, 255, cv2.THRESH_BINARY)[1]

    ########################
    # bounding box from bw
    ########################
    ret,thresh = cv2.threshold(bw,50,255,1)

    # Remove some small noise if any.
    dilate = cv2.dilate(thresh,None)
    erode = cv2.erode(dilate,None)
    cv2.imshow('gray',erode)
    # Find contours with cv2.RETR_CCOMP
    contours,hierarchy = cv2.findContours(erode,cv2.RETR_CCOMP,cv2.CHAIN_APPROX_SIMPLE)

    loc = []
    for i,cnt in enumerate(contours):
        # Check if it is an external contour and its area is more than 100
        if hierarchy[0,i,3] == -1 and cv2.contourArea(cnt)>5000:
            x,y,w,h = cv2.boundingRect(cnt)
            loc.extend(cv2.boundingRect(cnt))
            cv2.rectangle(img,(x,y),(x+w,y+h),(0,255,0),2)

            

    if plot:
        plt.subplot(231),plt.imshow(orig,cmap = 'gray')
        plt.title('Original Image'), plt.xticks([]), plt.yticks([])

        plt.subplot(232),plt.imshow(dn,cmap = 'gray')
        plt.title('Highly Denoised Image'), plt.xticks([]), plt.yticks([])

        plt.subplot(233),plt.imshow(green,cmap = 'gray')
        plt.title('Mask Image'), plt.xticks([]), plt.yticks([])

        plt.subplot(234),plt.imshow(km,cmap = 'gray')
        plt.title('KM Image'), plt.xticks([]), plt.yticks([])

        plt.subplot(235),plt.imshow(bw,cmap = 'gray')
        plt.title('BW Image'), plt.xticks([]), plt.yticks([])

        plt.subplot(236),plt.imshow(img,cmap = 'gray')
        plt.title('Bounding Box'), plt.xticks([]), plt.yticks([])


        # plt.savefig('results/result_'+str(img_path.split('/')[1].split('.')[0])+'.png')
        plt.show()

    return loc

if __name__ == '__main__':
    import time
    img_path = 'imgs/1.JPG'
    st = time.time()
    print(get_bbox(img_path))
    print('time:',time.time()-st)

I have hardcoded for green leaves. Modify to your need and experiment.