Vision Transformer

Search before asking

• I have searched the Roboflow Notebooks issues and found no similar bug report.

Notebook name

GOOGLE COLLAB

Bug

I am getting the 'Key Error'='annotations'
And Type of Error: Detections._Getitem not supported for index of type <class 'numpy.ndarray'>....
Please Do guide me with the following error how can i come up with it.
Will really appreciate it.

Environment

Google Colab

Minimal Reproducible Example

import random
import cv2
import numpy as np
import pandas as pd

utils

categories = TEST_DATASET.coco.cats
id2label = {k: v['name'] for k,v in categories.items()}
box_annotator = sv.BoxAnnotator()

select random image

image_ids = TEST_DATASET.coco.getImgIds()
image_id = random.choice(image_ids)
print('Image #{}'.format(image_id))

load image and annotatons

image = TEST_DATASET.coco.loadImgs(image_id)[0]
annotations = TEST_DATASET.coco.imgToAnns[image_id]
image_path = os.path.join(TEST_DATASET.root, image['file_name'])
image = cv2.imread(image_path)

annotate

detections = sv.Detections.from_coco_annotations(coco_annotation=annotations)
labels = [f"{id2label[class_id]}" for _, _, class_id, _ in detections]
frame = box_annotator.annotate(scene=image.copy(), detections=detections, labels=labels)

print('ground truth')
%matplotlib inline
sv.show_frame_in_notebook(frame, (16, 16))

inference

with torch.no_grad():

# load image and predict
inputs = image_processor(images=image, return_tensors='pt').to(DEVICE)
outputs = model(**inputs)

# post-process
target_sizes = torch.tensor([image.shape[:2]]).to(DEVICE)
results = image_processor.post_process_object_detection(
    outputs=outputs, 
    threshold=CONFIDENCE_TRESHOLD, 
    target_sizes=target_sizes
)[0]

annotate

detections_obj = sv.Detections.from_transformers(transformers_results=results).with_nms(threshold=0.5)
detections = detections_obj.get_data()
labels = [f"{id2label[class_id]} {confidence:.2f}" for _, confidence, class_id, _ in detections]
frame = box_annotator.annotate(scene=image.copy(), detections=detections_obj, labels=labels)

print(type(detections))
%matplotlib inline

sv.show_frame_in_notebook(frame, (16, 16))

Additional

No response

Are you willing to submit a PR?

• Yes I'd like to help by submitting a PR!

Start with the Google Colab minimal reproducible example, especially the Detections.from_coco_annotations, Detections.from_transformers, and get_data entry points. Re-run the ground-truth and inference annotation sections and compare the reported errors; done means the notebook completes both paths and displays the annotated output.