Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -5,7 +5,7 @@ import torch
|
|
5 |
import torchvision.transforms as transforms
|
6 |
|
7 |
from PIL import Image
|
8 |
-
from models.tag2text import tag2text_caption
|
9 |
|
10 |
import gradio as gr
|
11 |
|
@@ -17,56 +17,79 @@ normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
|
|
17 |
std=[0.229, 0.224, 0.225])
|
18 |
transform = transforms.Compose([transforms.Resize((image_size, image_size)),transforms.ToTensor(),normalize])
|
19 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
|
21 |
#######Swin Version
|
22 |
-
pretrained = '
|
23 |
|
24 |
-
|
25 |
|
26 |
-
|
27 |
-
|
28 |
|
29 |
|
30 |
-
def inference(raw_image, input_tag):
|
31 |
raw_image = raw_image.resize((image_size, image_size))
|
32 |
|
33 |
-
image = transform(raw_image).unsqueeze(0).to(device)
|
34 |
-
|
35 |
-
|
36 |
-
|
|
|
37 |
else:
|
38 |
-
|
39 |
-
|
40 |
-
|
|
|
|
|
|
|
|
|
|
|
41 |
|
42 |
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
|
|
|
|
|
|
50 |
|
51 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
52 |
|
|
|
53 |
|
54 |
-
|
|
|
55 |
|
56 |
-
|
57 |
|
58 |
-
title = "Tag2Text"
|
59 |
-
description = "Welcome to Tag2Text demo! (Supported by Fudan University, OPPO Research Institute, International Digital Economy Academy) <br/> Upload your image to get the <b>tags</b> and <b>caption</b> of the image. Optional: You can also input specified tags to get the corresponding caption."
|
60 |
|
61 |
-
article = "<p style='text-align: center'>
|
62 |
|
63 |
-
demo = gr.Interface(inference, inputs, outputs, title=title, description=description, article=article, examples=[
|
64 |
-
|
65 |
-
|
66 |
-
['images/
|
67 |
-
['images/
|
68 |
-
['images/
|
69 |
-
['images/1641173_2291260800.jpg',"none"],
|
70 |
])
|
71 |
|
72 |
-
demo.launch(enable_queue=True)
|
|
|
5 |
import torchvision.transforms as transforms
|
6 |
|
7 |
from PIL import Image
|
8 |
+
from models.tag2text import tag2text_caption, ram
|
9 |
|
10 |
import gradio as gr
|
11 |
|
|
|
17 |
std=[0.229, 0.224, 0.225])
|
18 |
transform = transforms.Compose([transforms.Resize((image_size, image_size)),transforms.ToTensor(),normalize])
|
19 |
|
20 |
+
#######Tag2Text Model
|
21 |
+
pretrained = '/home/notebook/data/group/huangxinyu/pretrain_model/tag2text/tag2text_swin_14m.pth'
|
22 |
+
|
23 |
+
model_tag2text = tag2text_caption(pretrained=pretrained, image_size=image_size, vit='swin_b' )
|
24 |
+
|
25 |
+
model_tag2text.eval()
|
26 |
+
model_tag2text = model_tag2text.to(device)
|
27 |
+
|
28 |
|
29 |
#######Swin Version
|
30 |
+
pretrained = '/home/notebook/code/personal/S9049611/tag2text-v2/output/pretrain_tag2text_large_v2_14m_large_v14/new_coco_ori_finetune_384_v5_epoch03/checkpoint_01.pth'
|
31 |
|
32 |
+
model_ram = ram(pretrained=pretrained, image_size=image_size, vit='swin_l' )
|
33 |
|
34 |
+
model_ram.eval()
|
35 |
+
model_ram = model_ram.to(device)
|
36 |
|
37 |
|
38 |
+
def inference(raw_image, model_n , input_tag):
|
39 |
raw_image = raw_image.resize((image_size, image_size))
|
40 |
|
41 |
+
image = transform(raw_image).unsqueeze(0).to(device)
|
42 |
+
if model_n == 'Recognize Anything Model':
|
43 |
+
model = model_ram
|
44 |
+
tags, tags_chinese = model.generate_tag(image)
|
45 |
+
return tags[0],tags_chinese[0], 'none'
|
46 |
else:
|
47 |
+
model = model_tag2text
|
48 |
+
model.threshold = 0.68
|
49 |
+
if input_tag == '' or input_tag == 'none' or input_tag == 'None':
|
50 |
+
input_tag_list = None
|
51 |
+
else:
|
52 |
+
input_tag_list = []
|
53 |
+
input_tag_list.append(input_tag.replace(',',' | '))
|
54 |
+
with torch.no_grad():
|
55 |
|
56 |
|
57 |
+
caption, tag_predict = model.generate(image,tag_input = input_tag_list,max_length = 50, return_tag_predict = True)
|
58 |
+
if input_tag_list == None:
|
59 |
+
tag_1 = tag_predict
|
60 |
+
tag_2 = ['none']
|
61 |
+
else:
|
62 |
+
_, tag_1 = model.generate(image,tag_input = None, max_length = 50, return_tag_predict = True)
|
63 |
+
tag_2 = tag_predict
|
64 |
+
|
65 |
+
return tag_1[0],'none',caption[0]
|
66 |
+
|
67 |
|
68 |
+
inputs = [
|
69 |
+
gr.inputs.Image(type='pil'),
|
70 |
+
gr.inputs.Radio(choices=['Recognize Anything Model',"Tag2Text Model"],
|
71 |
+
type="value",
|
72 |
+
default="Recognize Anything Model",
|
73 |
+
label="Model" ),
|
74 |
+
gr.inputs.Textbox(lines=2, label="User Specified Tags (Optional, Enter with commas, Currently only Tag2Text is supported)")
|
75 |
+
]
|
76 |
|
77 |
+
outputs = [gr.outputs.Textbox(label="Tags"),gr.outputs.Textbox(label="标签"), gr.outputs.Textbox(label="Caption (currently only Tag2Text is supported)")]
|
78 |
|
79 |
+
# title = "Recognize Anything Model"
|
80 |
+
title = "<font size='10'> Recognize Anything Model</font>"
|
81 |
|
82 |
+
description = "Welcome to the Recognize Anything Model (RAM) and Tag2Text Model demo! <li><b>Recognize Anything Model:</b> Upload your image to get the <b>English and Chinese outputs of the image tags</b>!</li><li><b>Tag2Text Model:</b> Upload your image to get the <b>tags</b> and <b>caption</b> of the image. Optional: You can also input specified tags to get the corresponding caption.</li> "
|
83 |
|
|
|
|
|
84 |
|
85 |
+
article = "<p style='text-align: center'>RAM and Tag2Text is training on open-source datasets, and we are persisting in refining and iterating upon it.<br/><a href='https://recognize-anything.github.io/' target='_blank'>Recognize Anything: A Strong Image Tagging Model</a> | <a href='https://https://tag2text.github.io/' target='_blank'>Tag2Text: Guiding Language-Image Model via Image Tagging</a> | <a href='https://github.com/xinyu1205/Tag2Text' target='_blank'>Github Repo</a></p>"
|
86 |
|
87 |
+
demo = gr.Interface(inference, inputs, outputs, title=title, description=description, article=article, examples=[
|
88 |
+
['images/demo1.jpg',"Recognize Anything Model","none"],
|
89 |
+
['images/demo2.jpg',"Recognize Anything Model","none"],
|
90 |
+
['images/demo4.jpg',"Recognize Anything Model","none"],
|
91 |
+
['images/demo4.jpg',"Tag2Text Model","power line"],
|
92 |
+
['images/demo4.jpg',"Tag2Text Model","track, train"] ,
|
|
|
93 |
])
|
94 |
|
95 |
+
demo.launch(enable_queue=True)
|