When body move in zigzag manner giving the wrong counting

Support YOLO
1 
import cv2

from ultralytics import solutions

cap = cv2.VideoCapture("/home/nasscom-gh-nwarch-ai/Documents/Siddharth/GSFC/GSFC_SIDD/Data_folder/172932574096445 (1).mp4")
assert cap.isOpened(), "Error reading video file"
w, h, fps = (int(cap.get(x)) for x in (cv2.CAP_PROP_FRAME_WIDTH, cv2.CAP_PROP_FRAME_HEIGHT, cv2.CAP_PROP_FPS))

# line or region points
#line_points = [(20, 400), (1580, 400)]
#best line we for our 01.mp4project
line_points = [(20, 500), (2400, 500)]
# Video writer
#video_writer = cv2.VideoWriter("object_counting_output.avi", cv2.VideoWriter_fourcc(*"mp4v"), fps, (w, h))

# Init Object Counter
counter = solutions.ObjectCounter(
    show=False,
    region=line_points,
    model="/home/nasscom-gh-nwarch-ai/Documents/Siddharth/GSFC/GSFC_SIDD/inferencing/GSFC_11_5_24.pt",
    
    #experiment
    line_width = 2,
    classes = [0],
    up_angle=90.0,
    down_angle = 270,
    angle_tolerance = 30.0
)
counter.model.imgsz=[640]
# counter.model.conf =0.65
# Process video
while cap.isOpened():
    success, im0 = cap.read()
    if not success:
        print("Video frame is empty or video processing has been successfully completed.")
        break
    im0 = counter.count(im0)
    #video_writer.write(im0)
    cv2.line(im0, (20, 600), (2400, 600), (0,0,255), 2)
    resized_frame = cv2.resize(im0, (960,640))
    #cv2.line(resized_frame, (20, 500), (1500, 500), (0,0,255), 2)
    cv2.imshow("Object Counting Output", resized_frame)
    if cv2.waitKey(int(1000 / fps)) & 0xFF == ord("q"):
        break

cap.release()
#video_writer.release()
cv2.destroyAllWindows()

suppose when i am working with some object like which move little bit in zigzag manner, so it give the wrong counting becuse when object little bit backward then it will move in the forward which result in the wrong counting

2

Hello there! :blush:

It sounds like you’re encountering an issue with zigzag movements affecting the accuracy of your object counting. This is a common challenge in object tracking and counting, especially when objects move back and forth across the counting line.

Here are a few suggestions to improve the accuracy:

  1. Adjust the Counting Logic: Consider implementing logic that accounts for the direction of movement. You can track the object’s trajectory and only count it when it crosses the line in a specific direction.

  2. Increase Angle Tolerance: You might want to experiment with the angle_tolerance parameter to see if a higher tolerance helps in reducing false counts due to zigzag movements.

  3. Use a Region Instead of a Line: Instead of a single line, define a region where the object must fully enter and exit before being counted. This can help in filtering out minor back-and-forth movements.

  4. Tracking Algorithm: Ensure that your tracking algorithm is robust enough to handle such movements. You might want to explore different tracking algorithms or fine-tune the existing one.

  5. Post-Processing: Implement post-processing logic to filter out false positives. For example, you could use a buffer to store recent object positions and determine if an object should be counted based on its overall movement pattern.

For more detailed guidance, you can check out the Ultralytics YOLO documentation and explore the Object Counting Guide.

Feel free to experiment with these suggestions and let us know how it goes! If you have any more questions, don’t hesitate to ask. Happy coding! :rocket:

3

In all likelihood, you’d need to modify the source code, probably around here to only update the counter when the ∆ of the direction changes from + → - or - → + for only the first time it crosses.

4

I got the point and implemented it the same way, and it’s working perfectly. Thank you so much for the help!