Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,97 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
import gradio as gr
|
3 |
+
import re
|
4 |
+
import cv2
|
5 |
+
from PIL import ImageDraw, Image
|
6 |
+
|
7 |
+
from transformers import AutoProcessor, PaliGemmaForConditionalGeneration
|
8 |
+
|
9 |
+
mix_model_id = "google/paligemma-3b-mix-224"
|
10 |
+
mix_model = PaliGemmaForConditionalGeneration.from_pretrained(mix_model_id)
|
11 |
+
mix_processor = AutoProcessor.from_pretrained(mix_model_id)
|
12 |
+
|
13 |
+
# Helper function to parse multiple <loc> tags and return a list of coordinate sets and labels
|
14 |
+
def parse_multiple_locations(decoded_output):
|
15 |
+
# Regex pattern to match four <locxxxx> tags and the label at the end (e.g., 'cat')
|
16 |
+
loc_pattern = r"<loc(\d{4})><loc(\d{4})><loc(\d{4})><loc(\d{4})>\s+(\w+)"
|
17 |
+
|
18 |
+
matches = re.findall(loc_pattern, decoded_output)
|
19 |
+
coords_and_labels = []
|
20 |
+
|
21 |
+
for match in matches:
|
22 |
+
# Extract the coordinates and label
|
23 |
+
y1 = int(match[0]) / 1000
|
24 |
+
x1 = int(match[1]) / 1000
|
25 |
+
y2 = int(match[2]) / 1000
|
26 |
+
x2 = int(match[3]) / 1000
|
27 |
+
label = match[4]
|
28 |
+
|
29 |
+
coords_and_labels.append({
|
30 |
+
'label': label,
|
31 |
+
'bbox': [y1, x1, y2, x2]
|
32 |
+
})
|
33 |
+
|
34 |
+
return coords_and_labels
|
35 |
+
|
36 |
+
# Helper function to draw bounding boxes and labels for all objects on the image
|
37 |
+
def draw_multiple_bounding_boxes(image, coords_and_labels):
|
38 |
+
draw = ImageDraw.Draw(image)
|
39 |
+
width, height = image.size
|
40 |
+
|
41 |
+
for obj in coords_and_labels:
|
42 |
+
# Extract the bounding box coordinates
|
43 |
+
y1, x1, y2, x2 = obj['bbox'][0] * height, obj['bbox'][1] * width, obj['bbox'][2] * height, obj['bbox'][3] * width
|
44 |
+
|
45 |
+
# Draw bounding box and label
|
46 |
+
draw.rectangle([x1, y1, x2, y2], outline="red", width=3)
|
47 |
+
draw.text((x1, y1), obj['label'], fill="red")
|
48 |
+
|
49 |
+
return image
|
50 |
+
|
51 |
+
# Define inference function
|
52 |
+
def process_image(image, prompt):
|
53 |
+
# Process the image and prompt using the processor
|
54 |
+
inputs = mix_processor(image.convert("RGB"), prompt, return_tensors="pt")
|
55 |
+
|
56 |
+
try:
|
57 |
+
# Generate output from the model
|
58 |
+
output = mix_model.generate(**inputs, max_new_tokens=100)
|
59 |
+
|
60 |
+
# Decode the output from the model
|
61 |
+
decoded_output = mix_processor.decode(output[0], skip_special_tokens=True)
|
62 |
+
|
63 |
+
# Extract bounding box coordinates and labels
|
64 |
+
coords_and_labels = parse_multiple_locations(decoded_output)
|
65 |
+
|
66 |
+
if coords_and_labels:
|
67 |
+
# Draw bounding boxes and labels on the image
|
68 |
+
image_with_boxes = draw_multiple_bounding_boxes(image, coords_and_labels)
|
69 |
+
|
70 |
+
# Prepare the coordinates and labels for the UI
|
71 |
+
labels_and_coords = "\n".join([f"Label: {obj['label']}, Coordinates: {obj['bbox']}" for obj in coords_and_labels])
|
72 |
+
|
73 |
+
# Return the modified image and the list of coordinates+labels
|
74 |
+
return image_with_boxes, labels_and_coords
|
75 |
+
else:
|
76 |
+
return "No bounding boxes detected."
|
77 |
+
|
78 |
+
except IndexError as e:
|
79 |
+
print(f"IndexError: {e}")
|
80 |
+
return "An error occurred during processing."
|
81 |
+
|
82 |
+
# Define the Gradio interface
|
83 |
+
inputs = [
|
84 |
+
gr.Image(type="pil"),
|
85 |
+
gr.Textbox(label="Prompt", placeholder="Enter your question")
|
86 |
+
]
|
87 |
+
outputs = [
|
88 |
+
gr.Image(label="Output Image with Bounding Boxes"),
|
89 |
+
gr.Textbox(label="Bounding Box Coordinates and Labels")
|
90 |
+
]
|
91 |
+
|
92 |
+
# Create the Gradio app
|
93 |
+
demo = gr.Interface(fn=process_image, inputs=inputs, outputs=outputs, title="Object Detection with Mix PaliGemma Model",
|
94 |
+
description="Upload an image and get object detections with bounding boxes and labels.")
|
95 |
+
|
96 |
+
# Launch the app
|
97 |
+
demo.launch(debug=True)
|