File size: 6,844 Bytes
6873ca3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ff9dd4b
6873ca3
 
 
 
 
 
 
 
 
 
 
ff9dd4b
6027eca
 
 
8790d1f
ff9dd4b
 
8790d1f
 
6873ca3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a80efbb
6873ca3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2b1561b
 
6873ca3
 
 
 
 
 
 
 
 
 
a80efbb
 
6873ca3
 
 
 
 
 
 
a80efbb
6873ca3
 
 
 
8cc4adb
 
 
ff9dd4b
ef33cf7
6027eca
6873ca3
 
 
 
ff9dd4b
6873ca3
 
 
 
ff9dd4b
6873ca3
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
import base64
import io
import cv2
import requests
import json
import gradio as gr
import os
from PIL import Image
import numpy as np
from PIL import ImageOps

# Accessing a specific environment variable
api_key = os.environ.get('devisionx')

# Checking if the environment variable exists
if not api_key:
    print("devisionx environment variable is not set.")
    exit()

# Define a function to call the API and get the results

def base64str_to_PILImage(base64str):
    base64_img_bytes = base64str.encode('utf-8')
    base64bytes = base64.b64decode(base64_img_bytes)
    bytesObj = io.BytesIO(base64bytes)
    return ImageOps.exif_transpose(Image.open(bytesObj))

def get_results(image, prompt,segment):
    threshold = 0.5
    
    # Convert the NumPy array to PIL image
    image = Image.fromarray(image)

    # Convert the image to base64 string
    with io.BytesIO() as output:
        image.save(output, format="JPEG")
        base64str = base64.b64encode(output.getvalue()).decode("utf-8")

    # Prepare the payload (Adjust this part according to the API requirements)
    #payload = json.dumps({"base64str": base64str, "classes": prompt})
    task_="0"
    if segment == "Segmentation":
        task_="1"
    payload =json.dumps({
        "base64str": base64str,
        "classes": prompt,
        "segment": task_  })
    # Prepare the payload (Adjust this part according to the API requirements)
    # Send the request to the API
    response = requests.put(api_key, data=payload)

    # Parse the JSON response
    data = response.json()
    print(response.status_code)
    print(data)

    # Access the values (Adjust this part according to the API response format)
    output_image_base64 = data['firstName']  # Assuming the API returns the output image as base64
    

    # Convert the output image from base64 to PIL and then to NumPy array
    output_image = base64str_to_PILImage(output_image_base64)
    output_image = np.array(output_image)

    return output_image

    
# Define the input components for Gradio (adding a new input for the prompt)
# image_input = gr.inputs.Image()
# text_input = gr.inputs.Textbox(label="Prompt")  # New input for the text prompt


# # Define the output components for Gradio (including both image and text)
# outputs = gr.Image(type="numpy", label="Output Image")

# Define the text description within an HTML <div> element
description_html = """
<!DOCTYPE html>
<html>
<head>
    <title>Tuba AI Auto-Annotation </title>
</head>
<body>
    <h1>Tuba AI Auto-Annotation 🚀</h1>
    <h2>Saving Time, Bounding Boxes & Polygons at a Time </h2>
    <h2>Introduction</h2>
    <p>Welcome to the world of DevisionX, where AI meets vision to revolutionize annotation. Our mission is to make computer vision accessible to all, and this README is your gateway to understanding how our auto-annotation model can change the way you work.</p>
    <h2>Meet Tuba.AI - Your Partner in Vision</h2>
    <h3>What is Tuba?</h3>
    <p>Tuba is the secret sauce behind DevisionX, your no-code/low-code companion for all things computer vision. It's your toolkit for labeling, training data, and deploying AI-vision applications faster and easier than ever before.</p>
    <ul>
        <li>No-Code/Low-Code: Say goodbye to complex coding. Tuba's user-friendly interface makes it accessible to everyone.</li>
        <li>Labeling Made Easy: Annotate your data effortlessly with Tuba's intuitive tools.</li>
        <li>Faster Deployment: Deploy your AI models with ease, whether you're building a standalone app or integrating within an existing one.</li>
        <li>State-of-the-Art Technology: Tuba is powered by the latest AI tech and follows production-ready standards.</li>
    </ul>
    <h2>The DevisionX Auto-Annotation</h2>
    <p>Our auto-annotation model is a game-changer. It takes input text and images, weaving them together to generate precise bounding boxes. This AI marvel comes with a plethora of benefits:</p>
    <ul>
        <li>Time Saver: Say goodbye to hours of manual annotation. Let our model do the heavy lifting.</li>
        <li>Annotation Formats: It speaks the language of YOLO and COCO, making it versatile for various projects.</li>
        <li>Human Assistance: While it's incredibly efficient, it also respects human creativity and can be your reliable assistant.</li>
    </ul>
    <h2>Let's Build Together</h2>
    <p>We are here to redefine the way you approach computer vision. Join us in this exciting journey, where AI meets creativity, and innovation knows no bounds.</p>
    <p>Get started today and be a part of the future of vision.</p>
</body>
</html>



"""
title = "autoannotation"

description = "This is a project description. It demonstrates how to use Gradio with an image and text input to interact with an API."

import os
examples = [
    ["traffic.jpg", 'person,car,traffic sign,traffic light', "Segmentation"],  # Example with "Segmentation" selected
    ["3000.jpeg", 'person,car,traffic sign,traffic light', "Detection"],  # Example with "Detection" selected
]



# Create a Blocks object and use it as a context manager
with gr.Blocks() as demo:
    gr.Markdown(
        """
        <div style="text-align: center;">
            <h1>Tuba Autoannotation Demo</h1>
            <h3>A prompt based controllable model for auto annotation (Detection and Segmentation) </h3>
            <h3>Saving Time, Bounding Boxes & Polygons at a Time </h3>
            Powered by <a href="https://Tuba.ai">Tuba</a>
        </div>
        """
        )
    # Define the input components and add them to the layout
    
    with gr.Row():
        image_input = gr.Image()
        output = gr.Image(type="numpy", label="Output Image")
        
    # Define the output component and add it to the layout
    with gr.Row():
        text_input = gr.Textbox(label="Prompt")

        # text_input = gr.inputs.Textbox(label="Prompt")
    with gr.Row():
        segment_checkbox = gr.Radio(["Segmentation", "Detection"], value="file",label="Select Detection or Segmentation",info="Select Segmentation to extract Polygons or Detection to extract only the bounding boxes of of the desired objects automatically")
        #segment_checkbox = gr.inputs.Checkbox(label="Segment", default=False)
    with gr.Row():
        button = gr.Button("Run")
        
    # Define the event listener that connects the input and output components and triggers the function
    button.click(fn=get_results, inputs=[image_input, text_input, segment_checkbox], outputs=output, api_name="get_results")
    # Add the description below the layout
    gr.Examples(
            fn=get_results,
            examples=examples,
            inputs=[image_input, text_input,segment_checkbox],
            outputs=[output]
        )
    gr.Markdown(description_html)
# Launch the app
demo.launch(share=False)