Spaces:
Sleeping
Sleeping
import gradio as gr | |
from transformers import DetrImageProcessor, DetrForObjectDetection | |
import torch | |
import supervision as sv | |
import json | |
import requests | |
from PIL import Image | |
import numpy as np | |
image_processor = DetrImageProcessor.from_pretrained("facebook/detr-resnet-50") | |
model = DetrForObjectDetection.from_pretrained("Guy2/AirportSec-150epoch") | |
id2label = {0: 'dangerous-items', 1: 'Gun', 2: 'Knife', 3: 'Pliers', 4: 'Scissors', 5: 'Wrench'} | |
def anylize(url): | |
image = Image.open(requests.get(url, stream=True).raw) | |
image = np.array(image) | |
with torch.no_grad(): | |
inputs = image_processor(images=image, return_tensors='pt') | |
outputs = model(**inputs) | |
target_sizes = torch.tensor([image.shape[:2]]) | |
results = image_processor.post_process_object_detection( | |
outputs=outputs, | |
threshold=0.8, | |
target_sizes=target_sizes | |
)[0] | |
# annotate | |
detections = sv.Detections.from_transformers(transformers_results=results).with_nms(threshold=0.5) | |
labels = [f"{id2label[class_id]} {confidence:.2f}" for _, _, confidence, class_id, _ in detections] | |
box_annotator = sv.BoxAnnotator() | |
frame = box_annotator.annotate(scene=image.copy(), detections=detections, labels=labels) | |
return frame | |
output = gr.components.Image(type="numpy", label="Output Image") | |
gr.Interface(fn = anylize, inputs="text", outputs=output).launch() | |