Hugging Face's logo

Spaces:
BAAI
/
SegGPT
Runtime error

App Files Community
1
SegGPT
File size: 7,314 Bytes 
3ee92ac
 
 
 
 
 
 
 
 
 
 
3dbd6ff
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3ee92ac
 
 
 
 
 
 
 
 
5f4cbf4
 
3ee92ac
 
 
d46ad28
 
 
 
3ee92ac
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3dbd6ff
ba4c64d
3dbd6ff
ba4c64d
3dbd6ff
 
 
 
 
3ee92ac
a2f30af
1602533
4b455f2
4c82dea
3ee92ac
 
 
 
 
 
72e357f
35ad4f6
4b455f2
35ad4f6
3ee92ac
 
 
 
 
 
3dbd6ff
 
 
 
 
 
 
 
 
 
 
b53d427
3dbd6ff
 
 
 
f9e9115
3dbd6ff
 
 
 
 
 
 
3ee92ac
 
 
 
 
e29e4d5
3ee92ac
 
 
 
3dbd6ff
3ee92ac
 
f057b7d
3ee92ac
 
 
 
 
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
 
# -*- coding: utf-8 -*-

import sys
import io
import requests
import json
import base64
from PIL import Image
import numpy as np
import gradio as gr


def inference_mask1_sam(prompt,
              img,
              img_):

    files = {
        "useSam" : 1,
        "pimage" : resizeImg(prompt["image"]),
        "pmask" : resizeImg(prompt["mask"]),
        "img" : resizeImg(img),
        "img_" : resizeImg(img_)
    }
    r = requests.post("http://120.92.79.209/painter/run", json = files)
    a = json.loads(r.text)

    res = []

    for i in range(len(a)):
        #out = Image.open(io.BytesIO(base64.b64decode(a[i])))
        #out = out.resize((224, 224))
        #res.append(np.uint8(np.array(out)))
        res.append(np.uint8(np.array(Image.open(io.BytesIO(base64.b64decode(a[i]))))))
    return res[1:] # remove prompt image

def inference_mask1(prompt,
              img,
              img_):
    files = {
        "pimage" : resizeImg(prompt["image"]),
        "pmask" : resizeImg(prompt["mask"]),
        "img" : resizeImg(img),
        "img_" : resizeImg(img_)
    }
    #r = requests.post("https://flagstudio.baai.ac.cn/painter/run", json = files)
    r = requests.post("http://120.92.79.209/painter/run", json = files)
    a = json.loads(r.text)
    res = []
    for i in range(len(a)):
        #out = Image.open(io.BytesIO(base64.b64decode(a[i])))
        #out = out.resize((224, 224))
        #res.append(np.uint8(np.array(out)))
        res.append(np.uint8(np.array(Image.open(io.BytesIO(base64.b64decode(a[i]))))))
    return res

def resizeImg(img):
    res, hres = 448, 448
    img = Image.fromarray(img).convert("RGB")
    img = img.resize((res, hres))
    temp = io.BytesIO()
    img.save(temp, format="WEBP")
    return base64.b64encode(temp.getvalue()).decode('ascii')

def inference_mask_cat(
              prompt,
              img,
              img_,
              ):
    output_list = [img, img_]
    return output_list


# define app features and run

examples = [
            ['./images/hmbb_1.jpg', './images/hmbb_2.jpg', './images/hmbb_3.jpg'],
            ['./images/rainbow_1.jpg', './images/rainbow_2.jpg', './images/rainbow_3.jpg'],
            ['./images/earth_1.jpg', './images/earth_2.jpg', './images/earth_3.jpg'],
            ['./images/obj_1.jpg', './images/obj_2.jpg', './images/obj_3.jpg'],
            ['./images/ydt_2.jpg', './images/ydt_1.jpg', './images/ydt_3.jpg'],
           ]

examples_sam = [
            ['./images/nc_1.jpg', './images/nc_2.jpg', './images/nc_3.jpg'],
            ['./images/street_1.jpg', './images/street_2.jpg', './images/street_3.jpg'],
            ['./images/hmbb_1.jpg', './images/hmbb_2.jpg', './images/hmbb_3.jpg'],
            ['./images/earth_1.jpg', './images/earth_2.jpg', './images/earth_3.jpg'],
            ['./images/ydt_2.jpg', './images/ydt_1.jpg', './images/ydt_3.jpg'],
           ]


demo_mask = gr.Interface(fn=inference_mask1, 
                   inputs=[gr.ImageMask(brush_radius=8, label="prompt (提示图)"), gr.Image(label="img1 (测试图1)"), gr.Image(label="img2 (测试图2)")], 
                    #outputs=[gr.Image(shape=(448, 448), label="output1 (输出图1)"), gr.Image(shape=(448, 448), label="output2 (输出图2)")],
                    outputs=[gr.Image(label="output1 (输出图1)").style(height=256, width=256), gr.Image(label="output2 (输出图2)").style(height=256, width=256)],
                    #outputs=gr.Gallery(label="outputs (输出图)"),
                    examples=examples,
                    #title="SegGPT for Any Segmentation<br>(Painter Inside)",
                    description="<p> \
                    Choose an example below &#128293; &#128293;  &#128293; <br>\
                    Or, upload by yourself: <br>\
                    1. Upload images to be tested to 'img1' and/or 'img2'. <br>2. Upload a prompt image to 'prompt' and draw a mask.  <br>\
                            <br> \
                            💎 The more accurate you annotate, the more accurate the model predicts. <br>\
                            💎 Examples below were never trained and are randomly selected for testing in the wild. <br>\
                            💎 Current UI interface only unleashes a small part of the capabilities of SegGPT, i.e., 1-shot case. \
</p>",
                   cache_examples=False,
                   allow_flagging="never",
                   )



demo_mask_sam = gr.Interface(fn=inference_mask1_sam,
                   inputs=[gr.ImageMask(brush_radius=4, label="prompt (提示图)"), gr.Image(label="img1 (测试图1)"), gr.Image(label="img2 (测试图2)")],
                    #outputs=[gr.Image(shape=(448, 448), label="output1 (输出图1)"), gr.Image(shape=(448, 448), label="output2 (输出图2)")],
                    # outputs=[gr.Image(label="output1 (输出图1)").style(height=256, width=256), gr.Image(label="output2 (输出图2)").style(height=256, width=256)],
                    #outputs=gr.Gallery(label="outputs (输出图)"),
                    outputs=[gr.Image(label="SAM output (mask)").style(height=256, width=256),gr.Image(label="output1 (输出图1)").style(height=256, width=256), gr.Image(label="output2 (输出图2)").style(height=256, width=256)],
                    # outputs=[gr.Image(label="output3 (输出图1)").style(height=256, width=256), gr.Image(label="output4 (输出图2)").style(height=256, width=256)],
                    examples=examples_sam,
                    #title="SegGPT for Any Segmentation<br>(Painter Inside)",
                    description="<p> \
                    <strong>SAM+SegGPT: One touch for segmentation in all images or videos.<strong> <br>\
                    Choose an example below &#128293; &#128293;  &#128293; <br>\
                    Or, upload by yourself: <br>\
                    1. Upload images to be tested to 'img1' and 'img2'. <br>2. Upload a prompt image to 'prompt' and draw <strong>a point or line on the target<strong>.  <br>\
                            <br> \
                            💎 SAM segments the target with any point or scribble, then SegGPT segments all other images. <br>\
                            💎 Examples below were never trained and are randomly selected for testing in the wild. <br>\
                            💎 Current UI interface only unleashes a small part of the capabilities of SegGPT, i.e., 1-shot case. \
</p>",
                   cache_examples=False,
                   allow_flagging="never",
                   )

title = "SegGPT: Segmenting Everything In Context<br> \
<div align='center'> \
<h2><a href='https://arxiv.org/abs/2304.03284' target='_blank' rel='noopener'>[paper]</a> \
<a href='https://github.com/baaivision/Painter' target='_blank' rel='noopener'>[code]</a></h2> \
<br> \
<image src='file/rainbow2.gif' width='720px' /> \
<h2>SegGPT performs arbitrary segmentation tasks in images or videos via in-context inference, such as object instance, stuff, part, contour, and text, with only one single model.</h2> \
</div> \
"

demo = gr.TabbedInterface([demo_mask_sam, demo_mask], ['SAM+SegGPT (一触百通)', 'General 1-shot'], title=title)

#demo.launch(share=True, auth=("baai", "vision"))
demo.launch(enable_queue=False)
#demo.launch(server_name="0.0.0.0", server_port=34311)
# -