Darknet YoloV4 empty prediction

Question

I'm trying to detect objects in an image in python using YoloV4 and Darknet. The problem is that it isn't detecting any objects in the image. This is my code:

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

netMain = darknet.load_net_custom(configPath.encode("ascii"), weightPath.encode("ascii"), 0, 1)
metaMain = darknet.load_meta(metaPath.encode("ascii"))
altNames = None
try:
    with open(metaPath) as metaFH:
        metaContents = metaFH.read()
        import re
        match = re.search("names *= *(.*)$", metaContents, re.IGNORECASE | re.MULTILINE)
        if match:
            result = match.group(1)
        else:
            result = None
        try:
            if os.path.exists(result):
                with open(result) as namesFH:
                     namesList = namesFH.read().strip().split("\n")
                     altNames = [x.strip() for x in namesList]
        except TypeError:
            pass
except Exception:
    pass

frame_width = darknet.network_width(netMain)
frame_height = darknet.network_height(netMain)

frame = cv2.imread("b.jpg")

darknet_image = darknet.make_image(frame_width, frame_height, 3)
darknet.copy_image_from_bytes(darknet_image, frame.tobytes())
detections = darknet.detect_image(netMain, metaMain, darknet_image, thresh=0.25)
print(detections) #returns []

The program prints "[]". The image ("b.jpg") is 764x756.

score 0 · Answer 1 · answered Feb 28 '21 at 06:22

0

You need to resize the image with proper size.

frame = cv2.resize(cv2.imread("b.jpg"), (frame_width, frame_height))

Make sure, the image has the objects you're looking for.

Before using the python script, use the darknet test to ensure the model can detect objects.

answered Feb 28 '21 at 06:22

Zabir Al Nazi

10,298
4
33
60

I tested the image with darknet test from cmd and it recognized two persons, so the image has the objects. I also tried resizing the image with your code but this time nothing happens, it never returns from darknet.detect_image() – Francisco Linan Feb 28 '21 at 06:50
In that case, can you try with the updated API from AlexeyAB repo? Or, you can try with this model server https://github.com/zabir-nabil/darknet-fastapi-modelserver – Zabir Al Nazi Feb 28 '21 at 07:13
What do you mean by the updated API? I'm using darknet from AlexeyAB (I compiled it myself) and I tried the cmd command "darknet.exe detector test cfg/coco.data cfg/yolov4.cfg yolov4.weights ", it works fine. – Francisco Linan Feb 28 '21 at 19:00

Darknet YoloV4 empty prediction

1 Answers1