3

I've been using the image AI library to create a simple object detection system using python, which currently records video using the webcam and as far as I can tell detects objects live, once the program stops, a video file is created with object markers overlayed on the video. The working code:

from imageai.Detection import VideoObjectDetection
import os
import cv2

camera = cv2.VideoCapture(0)

current_directory = os.getcwd()

detector = VideoObjectDetection()

detector.setModelTypeAsYOLOv3()

detector.setModelPath(r"C:\Users\Luca\Desktop\image\yolo.h5")
detector.loadModel()

detections = detector.detectObjectsFromVideo(
camera_input=camera,
output_file_path = r"C:\Users\Luca\Desktop\image\footage_detected", frames_per_second=20, log_progress=True)

for eachObject in detections:
    print(
         eachObject["name"] , " : ",
         eachObject["percentage_probability"], " : ",
         eachObject["box_points"] )
    print("--------------------------------")
vid.release()
cv2.destroyAllWindows()

I would like the program to display each frame on the display as it is being recorded with the object markers overlayed on the live feed as it detects them. I am only able to get the live camera feed directly from opencv without the object markers.

I attempted to use opencv and imageai to take each frame, use the detectObjectsFromImage method to process each frame and then using the output of the method with the opencv imshow method. However it gives the error:

ValueError: "Ensure you specified correct input image, input type, output type and/or output image path"

Here is the code where I attempt to do this:

from imageai.Detection import ObjectDetection
import os
import cv2

current_directory = os.getcwd()

camera = cv2.VideoCapture(0)


detector = ObjectDetection()

detector.setModelTypeAsYOLOv3()

detector.setModelPath(r"C:\Users\Luca\Desktop\image\yolo.h5")
detector.loadModel()
while True:
    ret, frame = camera.read()

    returned_image, detections = detector.detectObjectsFromImage(
    input_image = (frame),
    output_type = "array"
    )
    for eachObject in detections:
        print(
             eachObject["name"] , " : ",
             eachObject["percentage_probability"], " : ",
             eachObject["box_points"] )
        print("--------------------------------")

    cv2.imshow('frame', returned_image)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

Please let if this is possible, and if so how can it be done using the ImageAI library? Thank you in advance.

Edit: here is the complete traceback.

Traceback (most recent call last):
  File "C:\Users\Luca\anaconda3\envs\objectreconition\lib\site-packages\imageai\Detection\__init__.py", line 285, in detectObjectsFromImage
    input_image = cv2.imread(input_image)
TypeError: Can't convert object to 'str' for 'filename'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "C:\Users\Luca\Desktop\image\main 2.py", line 22, in <module>
    output_type = "array"
  File "C:\Users\Luca\anaconda3\envs\objectreconition\lib\site-packages\imageai\Detection\__init__.py", line 392, in detectObjectsFromImage
    "Ensure you specified correct input image, input type, output type and/or output image path ")
ValueError: Ensure you specified correct input image, input type, output type and/or output image path
Christoph Rackwitz
  • 11,317
  • 4
  • 27
  • 36
jerico
  • 31
  • 4
  • complete traceback please. -- and where's the error checking that is needed when using VideoCapture? don't follow tutorials that don't show this properly – Christoph Rackwitz Jun 15 '22 at 11:43
  • sorry could you please clarify - would you like me to post the whole error? – jerico Jun 15 '22 at 11:45
  • yes always the complete traceback. [edit] your question to add it. -- next you'll have to debug your code, i.e. figure out what values are in your variables, vs. what you *think* is in there. – Christoph Rackwitz Jun 15 '22 at 11:46
  • @jerico The error occurs when you initiate `detectObjectsFromImage()`. `input_image` has to be an image file not an array `frame`. That's why the error "Ensure you specified correct input image, input type, ...." – Jeru Luke Jun 15 '22 at 11:53
  • As per the [documentation](https://imageai.readthedocs.io/en/latest/detection/index.html) you cannot pass an array as input, only an image file – Jeru Luke Jun 15 '22 at 11:55
  • is there a way to somehow use it as an input? – jerico Jun 15 '22 at 12:15
  • @ChristophRackwitz I've added full trackback – jerico Jun 15 '22 at 12:21
  • @jerico On [this page](https://imageai.readthedocs.io/en/latest/video/index.html) there is a snippet that would show detected images once per second – Jeru Luke Jun 15 '22 at 12:26
  • your `imread` is wrong. you passed something other than a string. or rather, you use imageai in some way it's not designed for. it expects `input_image` to be a path. imageai documentation (very sparse!) indicates that you need to set `input_type`. their api design and documentation is awful. -- imageAI hasn't received any development for over a year. perhaps it's dead... it doesn't have many stars and forks on github either, so it seems to be unpopular. perhaps you might want to look for some other library. – Christoph Rackwitz Jun 15 '22 at 13:03

0 Answers0