GroundingDINO autodistill inference troubleshooting

[SCAlex91]

Search before asking

• I have searched the Autodistill issues and found no similar feature requests.

Question

I am trying to create some annotations using GroundingDINO using "inference" and "autodistill" libraries, and visualize the detections and annotations that were applied to the images. I can't use plot function because I am using the terminal to access the arch linux node, and I am trying to write the annotations to each image using "cv2". The following code works, but when I inspect the images, the detections are not showing at all on the images. Any ideas on what might be going wrong?

`from autodistill_grounding_dino import GroundingDINO
from autodistill.detection import CaptionOntology
from autodistill.utils import plot
import supervision as sv
import cv2
import os

images = ["../Images/batch_12/000062.jpg", "../Images/batch_1/000014.jpg", "../Images/batch_12/000061.jpg", "../Images/batch_12/000043.jpg", "../Images/batch_12/000084.jpg"]

for image in images:

image_dir, image_file = os.path.split(image)
image_filename, image_extension = os.path.splitext(image_file)

image = cv2.imread(image)

ontology = CaptionOntology(
    {
        "garbage": "garbage",
    }
)

model = GroundingDINO(ontology=ontology)

result = model.predict(image)

detections = sv.Detections.from_inference(result)

bounding_box_annotator = sv.BoundingBoxAnnotator(thickness=8, color=sv.ColorPalette.DEFAULT.by_idx(12))
label_annotator = sv.LabelAnnotator(text_thickness=3, text_scale=1.5, text_color=sv.Color.WHITE, color=sv.Color.ROBOFLOW)

labels = [
f"{model.ontology.classes()[class_id]} {confidence:0.2f}"
for _, _, confidence, class_id, _ in detections
]

annotated_image = bounding_box_annotator.annotate(
	scene=image.copy(), detections=detections
)

annotated_image = label_annotator.annotate(
scene=annotated_image, detections=detections, labels=labels
)

cv2.imwrite(f'./predictions/{image_filename}_groundingDINO_autodistill.jpg', annotated_image)`

Additional

No response

[capjamesg]

Hello there! If you run print(detections), what do you get?