AbdulManan093
commited on
Update app.py
Browse files
app.py
CHANGED
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from transformers import DetrImageProcessor, DetrForObjectDetection
|
2 |
+
import torch
|
3 |
+
from PIL import Image
|
4 |
+
import matplotlib.pyplot as plt
|
5 |
+
import matplotlib.patches as patches
|
6 |
+
import gradio as gr
|
7 |
+
import io
|
8 |
+
|
9 |
+
# Load the processor and model
|
10 |
+
processor = DetrImageProcessor.from_pretrained("facebook/detr-resnet-50", revision="no_timm")
|
11 |
+
model = DetrForObjectDetection.from_pretrained("facebook/detr-resnet-50", revision="no_timm")
|
12 |
+
|
13 |
+
def detect_and_display_image(image):
|
14 |
+
# Ensure image is in PIL format
|
15 |
+
if isinstance(image, bytes):
|
16 |
+
image = Image.open(io.BytesIO(image))
|
17 |
+
elif isinstance(image, str):
|
18 |
+
image = Image.open(image)
|
19 |
+
|
20 |
+
# Process the image
|
21 |
+
inputs = processor(images=image, return_tensors="pt")
|
22 |
+
|
23 |
+
# Perform object detection
|
24 |
+
outputs = model(**inputs)
|
25 |
+
|
26 |
+
# Convert outputs to COCO API format
|
27 |
+
target_sizes = torch.tensor([image.size[::-1]])
|
28 |
+
results = processor.post_process_object_detection(outputs, target_sizes=target_sizes, threshold=0.9)[0]
|
29 |
+
|
30 |
+
# Create a figure and axis for visualization
|
31 |
+
fig, ax = plt.subplots(1, figsize=(12, 9))
|
32 |
+
ax.imshow(image)
|
33 |
+
|
34 |
+
# Add bounding boxes and labels to the image
|
35 |
+
for score, label, box in zip(results["scores"], results["labels"], results["boxes"]):
|
36 |
+
box = [round(i, 2) for i in box.tolist()]
|
37 |
+
# Create a Rectangle patch
|
38 |
+
rect = patches.Rectangle(
|
39 |
+
(box[0], box[1]),
|
40 |
+
box[2] - box[0],
|
41 |
+
box[3] - box[1],
|
42 |
+
linewidth=2,
|
43 |
+
edgecolor='red',
|
44 |
+
facecolor='none'
|
45 |
+
)
|
46 |
+
# Add the patch to the Axes
|
47 |
+
ax.add_patch(rect)
|
48 |
+
# Add label and confidence score
|
49 |
+
plt.text(
|
50 |
+
box[0], box[1],
|
51 |
+
f'{model.config.id2label[label.item()]}: {round(score.item(), 3)}',
|
52 |
+
color='red',
|
53 |
+
fontsize=12,
|
54 |
+
bbox=dict(facecolor='yellow', alpha=0.5)
|
55 |
+
)
|
56 |
+
|
57 |
+
plt.axis('off') # Hide the axes
|
58 |
+
|
59 |
+
# Save the figure to a BytesIO object and return it
|
60 |
+
buf = io.BytesIO()
|
61 |
+
plt.savefig(buf, format='png')
|
62 |
+
buf.seek(0)
|
63 |
+
return Image.open(buf)
|
64 |
+
|
65 |
+
# Create a Gradio interface
|
66 |
+
iface = gr.Interface(
|
67 |
+
fn=detect_and_display_image,
|
68 |
+
inputs=gr.Image(type="pil"),
|
69 |
+
outputs=gr.Image(type="pil"),
|
70 |
+
title="Object Detection with DETR",
|
71 |
+
description="Upload an image to detect objects using the DETR model.",
|
72 |
+
live=True
|
73 |
+
)
|
74 |
+
|
75 |
+
# Launch the Gradio app
|
76 |
+
iface.launch()
|
77 |
+
|
78 |
+
|