Help with coding - Detecting an Object in a specific area

Python
1

Hello,
I would like to make an MXB headlamp for my motorbike with the help of Python and OpenCV, so practically to detect persons, cars, motorcycles, busses and so on…on the road, and then turn of there the light stripe in that region so that it gets blended out.

With the help of youtube videos i have managed to get this far:

image
image1608×952 73 KB

But i don’t know how to do the following:

in the segments from 1 to 12, i have to detect the persons, cars, motorcycles, and then i should be able to define an action for that region.

Example1: the person is detected in the region MXB1:MXB9(picture above) → all these stripes should go off, all others should remain on.

Example2: a car is being detected in the region of MXB2, then only the mxb2 should go off, and all others should remain on, so that you can define for which classIDS the action should be performed in an designated area.

could you help me with a hint how i can perform this action ?
I don’t know if you have understood what i want to achieve, but i hope so.

Thank you
Flo

my code is here:

import cv2
import numpy as np
from cv2 import dnn_DetectionModel
import time
#import tensorflow as tf
#import argparse
#import sys


thres = 0.45  # Threshold to detect object

cap = cv2.VideoCapture(0)
cap.set(3, 1280)
cap.set(4, 720)
cap.set(10, 70)

# Set up camera constants
IM_WIDTH = 1280
IM_HEIGHT = 720



classNames = []
classFile = 'coco.names'
with open(classFile,'rt') as f:
    classNames = f.read().rstrip('\n').split('\n')

configPath = 'ssd_mobilenet_v3_large_coco_2020_01_14.pbtxt'
weightsPath = 'frozen_inference_graph.pb'

net: dnn_DetectionModel = cv2.dnn_DetectionModel(weightsPath, configPath)
net.setInputSize(320, 320)
net.setInputScale(1.0 / 127.5)
net.setInputMean((127.5, 127.5, 127.5))
net.setInputSwapRB(True)



# Define MXB1 box coordinates (top left and bottom right)
TL_mx1 = (int(IM_WIDTH * 0.40), int(IM_HEIGHT * 0.01))
BR_mx1 = (int(IM_WIDTH * 0.44), int(IM_HEIGHT * 0.7))

# Define MXB2 box coordinates (top left and bottom right)
TL_mx2 = (int(IM_WIDTH * 0.44), int(IM_HEIGHT * 0.01))
BR_mx2 = (int(IM_WIDTH * 0.48), int(IM_HEIGHT * 0.7))

# Define MXB3 box coordinates (top left and bottom right)
TL_mx3 = (int(IM_WIDTH * 0.48), int(IM_HEIGHT * 0.01))
BR_mx3 = (int(IM_WIDTH * 0.52), int(IM_HEIGHT * 0.7))

# Define MXB4 box coordinates (top left and bottom right)
TL_mx4 = (int(IM_WIDTH * 0.52), int(IM_HEIGHT * 0.01))
BR_mx4 = (int(IM_WIDTH * 0.56), int(IM_HEIGHT * 0.7))

# Define MXB5 box coordinates (top left and bottom right)
TL_mx5 = (int(IM_WIDTH * 0.56), int(IM_HEIGHT * 0.01))
BR_mx5 = (int(IM_WIDTH * 0.60), int(IM_HEIGHT * 0.7))

# Define MXB6 box coordinates (top left and bottom right)
TL_mx6 = (int(IM_WIDTH * 0.60), int(IM_HEIGHT * 0.01))
BR_mx6 = (int(IM_WIDTH * 0.64), int(IM_HEIGHT * 0.7))

# Define MXB7 box coordinates (top left and bottom right)
TL_mx7 = (int(IM_WIDTH * 0.64), int(IM_HEIGHT * 0.01))
BR_mx7 = (int(IM_WIDTH * 0.68), int(IM_HEIGHT * 0.7))

# Define MXB8 box coordinates (top left and bottom right)
TL_mx8 = (int(IM_WIDTH * 0.68), int(IM_HEIGHT * 0.01))
BR_mx8 = (int(IM_WIDTH * 0.72), int(IM_HEIGHT * 0.7))

# Define MXB9 box coordinates (top left and bottom right)
TL_mx9 = (int(IM_WIDTH * 0.72), int(IM_HEIGHT * 0.01))
BR_mx9 = (int(IM_WIDTH * 0.76), int(IM_HEIGHT * 0.7))

# Define MXB10 box coordinates (top left and bottom right)
TL_mx10 = (int(IM_WIDTH * 0.76), int(IM_HEIGHT * 0.01))
BR_mx10 = (int(IM_WIDTH * 0.80), int(IM_HEIGHT * 0.7))

# Define MXB11 box coordinates (top left and bottom right)
TL_mx11 = (int(IM_WIDTH * 0.80), int(IM_HEIGHT * 0.01))
BR_mx11 = (int(IM_WIDTH * 0.84), int(IM_HEIGHT * 0.7))

# Define MXB12 box coordinates (top left and bottom right)
TL_mx12 = (int(IM_WIDTH * 0.84), int(IM_HEIGHT * 0.01))
BR_mx12 = (int(IM_WIDTH * 0.95), int(IM_HEIGHT * 0.7))

# Initialize control variables used for person detector

detected_inside = False
detected_outside = False

inside_counter = 0
outside_counter = 0

pause = 0
pause_counter = 0





font = cv2.FONT_HERSHEY_SIMPLEX

pTime = 0
while True:
    success, img = cap.read()
    confs: object
    classIds, confs, bbox = net.detect(img, confThreshold=0.5)
    bbox = list(bbox)
    confs = list(np.array(confs).reshape(1,-1)[0])
    confs = list(map(float,confs))

    #print(type(confs[0]))
    print(confs)


    indices = cv2.dnn.NMSBoxes(bbox, confs, thres, nms_threshold=0.2)
    print(indices, bbox)

    for i in indices:
        #i = i[0]
        box = bbox[i]
        x,y,w,h = box[0],box[1],box[2],box[3]


        cv2.rectangle(img, TL_mx1, BR_mx1, (255, 20, 20), 1)
        cv2.putText(img, "1", (TL_mx1[0] + 10, TL_mx1[1] + 30), font, 1, (255, 20, 255), 3,
                    cv2.LINE_AA)

        cv2.rectangle(img, TL_mx2, BR_mx2, (255, 20, 20), 1)
        cv2.putText(img, "2", (TL_mx2[0] + 10, TL_mx2[1] + 30), font, 1, (255, 20, 255), 3,
                    cv2.LINE_AA)

        cv2.rectangle(img, TL_mx3, BR_mx3, (255, 20, 20), 1)
        cv2.putText(img, "3", (TL_mx3[0] + 10, TL_mx3[1] + 30), font, 1, (255, 20, 255), 3,
                    cv2.LINE_AA)

        cv2.rectangle(img, TL_mx4, BR_mx4, (255, 20, 20), 1)
        cv2.putText(img, "4", (TL_mx4[0] + 10, TL_mx4[1] + 30), font, 1, (255, 20, 255), 3,
                    cv2.LINE_AA)

        cv2.rectangle(img, TL_mx5, BR_mx5, (255, 20, 20), 1)
        cv2.putText(img, "5", (TL_mx5[0] + 10, TL_mx5[1] + 30), font, 1, (255, 20, 255), 3,
                    cv2.LINE_AA)

        cv2.rectangle(img, TL_mx6, BR_mx6, (255, 20, 20), 1)
        cv2.putText(img, "6", (TL_mx6[0] + 10, TL_mx6[1] + 30), font, 1, (255, 20, 255), 3,
                    cv2.LINE_AA)

        cv2.rectangle(img, TL_mx7, BR_mx7, (255, 20, 20), 1)
        cv2.putText(img, "7", (TL_mx7[0] + 10, TL_mx7[1] + 30), font, 1, (255, 20, 255), 3,
                    cv2.LINE_AA)

        cv2.rectangle(img, TL_mx8, BR_mx8, (255, 20, 20), 1)
        cv2.putText(img, "8", (TL_mx8[0] + 10, TL_mx8[1] + 30), font, 1, (255, 20, 255), 3,
                    cv2.LINE_AA)

        cv2.rectangle(img, TL_mx9, BR_mx9, (255, 20, 20), 1)
        cv2.putText(img, "9", (TL_mx9[0] + 10, TL_mx9[1] + 30), font, 1, (255, 20, 255), 3,
                    cv2.LINE_AA)

        cv2.rectangle(img, TL_mx10, BR_mx10, (255, 20, 20), 1)
        cv2.putText(img, "10", (TL_mx10[0] + 10, TL_mx10[1] + 30), font, 1, (255, 20, 255), 3,
                    cv2.LINE_AA)

        cv2.rectangle(img, TL_mx11, BR_mx11, (255, 20, 20), 1)
        cv2.putText(img, "11", (TL_mx11[0] + 10, TL_mx11[1] + 30), font, 1, (255, 20, 255), 3,
                    cv2.LINE_AA)

        cv2.rectangle(img, TL_mx12, BR_mx12, (255, 20, 20), 1)
        cv2.putText(img, "12", (TL_mx12[0] + 10, TL_mx12[1] + 30), font, 1, (255, 20, 255), 3,
                    cv2.LINE_AA)




        cv2.rectangle(img, (x,y), (x+w,h+y), color=(0, 255, 0), thickness=2)
        cv2.putText(img,f'{classNames[classIds[i]-1].upper()} {int(confs[i]*100)}%', (x,y -10),
                    cv2.FONT_HERSHEY_COMPLEX, 0.6, (0, 255, 0), 2)

 #   if len(classIds) != 0:
 #       for classId, confidence, box in zip(classIds.flatten(), confs.flatten(), bbox):
  #              cv2.rectangle(img, box, color=(0, 255, 0), thickness=1),
   #             cv2.putText(img, classNames[classId-1].upper(), (box[0] + 10, box[1] + 10),
    #                        cv2.FONT_HERSHEY_COMPLEX, 1, (0, 255, 0), 2)
     #           cv2.putText(img, str(round(confidence * 100, 2)), (box[0] + 200, box[1] + 30),
      #                      cv2.FONT_HERSHEY_COMPLEX, 1, (0, 255, 0), 2)

#Here the FPS are generated
    cTime = time.time()
    fps = 1 / (cTime - pTime)
    pTime = cTime
    cv2.putText(img, f'FPS: {int(fps)}', (20,70), cv2.FONT_HERSHEY_SIMPLEX, 1, (0,255,0),2)

    cv2.imshow("Output", img)
    cv2.waitKey(1)
2

unfortunately, noone here will debug your ~200 lines “wall of code” …

can you explain, what this is for ? the purpose of the program ?

maybe you ned to rephrase the problem on a more technical level, like:

i'm looking for the intersection of an object detection with 12 predifined zones in the image

(is that it ?)

3

sounds like you basically need “rectangle-rectangle intersection” or just the intersection test without knowing the exact intersection rectangle.

that’s a simple geometric calculation and can be done per-axis, as a test for overlapping intervals.

and may I recommend that you use lists and loops? you have a ton of replicated code. I like the visual correspondence between things and code, but such code is not sustainable™. now’s a good time to collapse that, even if you just copy-pasted it during prototyping.

4

Hello @berak
thank you for your reply, that’s it ! this is actually what i want to achieve.
And as soon as an defined object intersects with those 12 predefined zones, an action should be performed(that action has to be defined for each zone).

I understand that noone will debug it, i just wanted to make clear what i have in my code.
I think i also need to make clear that i’m a complete beginner in coding…so yes. this is where i’m at.

Thank you for any hint.
Flo

5

ok, it’s sound simple, but for a complete beginner: you have just told me that i have to build a space ship !

“and may I recommend that you use lists and loops? you have a ton of replicated code. I like the visual correspondence between things and code, but such code is not sustainable™. now’s a good time to collapse that, even if you just copy-pasted it during prototyping.” → sorry once again, a complete beginner, i will search how to list & loop.

Thank you