Object of thype 'METADATA' has no len()

[phil100vol]

Hi,

I want to detect objects on my rtsp stream. Unfortunate the code ends in the following error:

File "/home/cctv/darknet/darknet.py", line 174, in detect_image
    do_nms_sort(detections, num, len(class_names), nms)
TypeError: object of type 'METADATA' has no len()

My minimal example looks like that:

# -*- coding: utf-8 -*-

import os
import cv2
import numpy as np
import time
import darknet

######################
sUser = "user"
sPassword = "1234"
sChannel = str(12)

configPath = "./cfg/yolov4.cfg"
weightPath = "./yolov4.weights"
metaPath = "./cfg/coco.data"

def cvDrawBoxes(detections, img):
    # Colored labels dictionary
    color_dict = {
        "person" : [0, 255, 255]
    }

    for detection in detections:
        x, y, w, h = detection[2][0],\
            detection[2][1],\
            detection[2][2],\
            detection[2][3]
        name_tag = str(detection[0].decode())
        for name_key, color_val in color_dict.items():
            if name_key == name_tag:
                color = color_val 
                xmin, ymin, xmax, ymax = convertBack(
                float(x), float(y), float(w), float(h))
                pt1 = (xmin, ymin)
                pt2 = (xmax, ymax)
                cv2.rectangle(img, pt1, pt2, color, 1)
                cv2.putText(img,
                            detection[0].decode() +
                            " [" + str(round(detection[1] * 100, 2)) + "]",
                            (pt1[0], pt1[1] - 5), cv2.FONT_HERSHEY_SIMPLEX, 0.5,
                            color, 2)
    return img

netMain = darknet.load_net_custom(configPath.encode("ascii"), weightPath.encode("ascii"), 0, 1)   
metaMain = darknet.load_meta(metaPath.encode("ascii"))

#convertion rtsp stream
os.environ["OPENCV_FFMPEG_CAPTURE_OPTIONS"] = "rtsp_transport;udp"

cap = cv2.VideoCapture(str('rtsp://')+sUser+str(':')+sPassword+str('@')+str('192.168.0.21')+str(':554/')+sChannel,cv2.CAP_FFMPEG)

_, frame = cap.read()
rows,cols,channels = frame.shape
darknet_image = darknet.make_image(cols,rows,3)

### LOOP ###

while True:

    _, frame = cap.read()
    frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB) 
    frame_resized = cv2.resize(frame_rgb,
                                   (cols, rows),
                                   interpolation=cv2.INTER_LINEAR)
    darknet.copy_image_from_bytes(darknet_image,frame_resized.tobytes()) 

    ###########
    detections = darknet.detect_image(netMain, metaMain, darknet_image, thresh=0.25)
    ###########

    image = cvDrawBoxes(detections, frame)

    cv2.imshow("image", frame)

    if cv2.waitKey(1) & 0xFF == ord("q"):
        break

cap.release
cv2.destroyAllWindows()`

Hope someone can give me a hand. Thanks

Phil

[eandert]

It seems you wrote this code for an older version of darknet.py prior to the changes made in 52c7314. The newest version of darknet.py has changed in some subtle ways. I suggest checking out https://github.com/AlexeyAB/darknet/blob/master/darknet_images.py for examples of how the newer version can be called.

You will need to change your code here:

netMain = darknet.load_net_custom(configPath.encode("ascii"), weightPath.encode("ascii"), 0, 1)
metaMain = darknet.load_meta(metaPath.encode("ascii"))

to be compatable with the newer version to something along these lines:

network, class_names, class_colors = darknet.load_network(configPath,  metaPath, weightPath, batch_size=1)

and then change your old call to detect_image: detections = darknet.detect_image(netMain, metaMain, darknet_image, thresh=0.25) to something more along these lines: detections = darknet.detect_image(network, class_names, darknet_image, thresh=0.25)

And finally within your cvDrawBoxes function:

for detection in detections:
        x, y, w, h = detection[2][0],\
            detection[2][1],\
            detection[2][2],\
            detection[2][3]
        name_tag = str(detection[0].decode())

Will need to be changed to something along the lines of:

for label, confidence, bbox in detections:
       x, y, w, h = (bbox[0],
             bbox[1],
             bbox[2],
             bbox[3])
       name_tag = label

and finally any reference to detection[1] in cvDrawBoxes should be changed to confidence

Hopefully I hit all of the major points here but I can't guarantee I caught every error. I recently had to update some older code as well and was having the same error so hope this helps. My code seems to be working after all of these fixes.

[sumanjob1988]

@eandert I got the same error and thanks to you, your solution works for me.

[tejask00]

netMain = darknet.load_net_custom(configPath.encode("ascii"), weightPath.encode("ascii"), 0, 1) metaMain = darknet.load_meta(metaPath.encode("ascii"))

do i have to change these lines with network, class_names, class_colors = darknet.load_network(configPath, metaPath, weightPath, batch_size=1) these?

[sfc9982]

also helpful for me, same issue

[Fadil-1]

@eandert YOU ROCK!. You saved me hours of headache. Thank you so much! And to the people who are not sure about @eandert answer's because he/she wrote "along these lines", try to do exactly what he/she said. It worked for me.

Namely replace each of netMain = darknet.load_net_custom(configPath.encode("ascii"), weightPath.encode("ascii"), 0, 1) and metaMain = darknet.load_meta(metaPath.encode("ascii")) with network, class_names, class_colors = darknet.load_network(configPath, metaPath, weightPath, batch_size=1)

detections = darknet.detect_image(netMain, metaMain, darknet_image, thresh=0.25) with detections = darknet.detect_image(network, class_names, darknet_image, thresh=0.25)

as well as the modifications in the for loop! Try it it works.

Thanks again @eandert