File size: 6,853 Bytes
b9288df
 
 
 
 
 
f8c1035
b9288df
 
 
 
 
 
1aa70ca
 
e337a53
b9288df
 
 
 
c6f3a23
b9288df
 
 
 
 
 
e337a53
b9288df
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1aa70ca
 
 
 
 
 
b9288df
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
16564f3
 
 
b9288df
 
 
 
 
 
 
16564f3
 
 
 
 
 
b9288df
 
 
 
 
 
 
 
 
 
 
1aa70ca
 
 
16564f3
1aa70ca
 
 
 
 
 
 
ec3f141
1aa70ca
 
 
 
 
b9288df
 
 
 
 
 
 
 
 
 
 
 
 
 
 
16564f3
 
b9288df
16564f3
b9288df
 
 
 
ef5f3c1
 
 
 
 
b9288df
 
 
 
16564f3
 
 
 
 
b9288df
 
 
 
 
 
ef5f3c1
 
 
 
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
166
167
168
169
170
171
172
173
174
175
176
177
178
import os
import numpy as np
import time
import random
import torch
import torchvision.transforms as transforms
import gradio as gr
import matplotlib.pyplot as plt

from models import get_model
from dotmap import DotMap
from PIL import Image

#os.environ['TERM'] = 'linux'
#os.environ['TERMINFO'] = '/etc/terminfo'

# args
args = DotMap()
args.deploy = 'vanilla'
args.arch = 'dino_small_patch16'
args.no_pretrain = True
args.resume = 'https://huggingface.co/hushell/pmf_dinosmall_lr1e-4/resolve/main/best_converted.pth'
args.api_key = 'AIzaSyAFkOGnXhy-2ZB0imDvNNqf2rHb98vR_qY'
args.cx = '06d75168141bc47f1'


# model
device = 'cpu' #torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = get_model(args)
model.to(device)
checkpoint = torch.hub.load_state_dict_from_url(args.resume, map_location='cpu')
model.load_state_dict(checkpoint['model'], strict=True)


# image transforms
def test_transform():
    def _convert_image_to_rgb(im):
        return im.convert('RGB')

    return transforms.Compose([
        transforms.Resize(256),
        transforms.CenterCrop(224),
        _convert_image_to_rgb,
        transforms.ToTensor(),
        transforms.Normalize(mean=[0.485, 0.456, 0.406],
                             std=[0.229, 0.224, 0.225]),
        ])

preprocess = test_transform()

@torch.no_grad()
def denormalize(x, mean, std):
    # 3, H, W
    t = x.clone()
    t.mul_(std).add_(mean)
    return torch.clamp(t, 0, 1)


# Google image search
from google_images_search import GoogleImagesSearch

class MyGIS(GoogleImagesSearch):
    def __enter__(self):
        return self
    def __exit__(self, exc_type, exc_val, exc_tb):
        return

# define search params
# option for commonly used search param are shown below for easy reference.
# For param marked with '##':
#   - Multiselect is currently not feasible. Choose ONE option only
#   - This param can also be omitted from _search_params if you do not wish to define any value
_search_params = {
    'q': '...',
    'num': 10,
    'fileType': 'png', #'jpg|gif|png',
    'rights': 'cc_publicdomain', #'cc_publicdomain|cc_attribute|cc_sharealike|cc_noncommercial|cc_nonderived',
    #'safe': 'active|high|medium|off|safeUndefined', ##
    'imgType': 'photo', #'clipart|face|lineart|stock|photo|animated|imgTypeUndefined', ##
    #'imgSize': 'huge|icon|large|medium|small|xlarge|xxlarge|imgSizeUndefined', ##
    #'imgDominantColor': 'black|blue|brown|gray|green|orange|pink|purple|red|teal|white|yellow|imgDominantColorUndefined', ##
    'imgColorType': 'color', #'color|gray|mono|trans|imgColorTypeUndefined' ##
}


# Gradio UI
def inference(query, labels, n_supp=10,
              file_type='png', rights='cc_publicdomain',
              image_type='photo', color_type='color'):
    '''
    query: PIL image
    labels: list of class names
    '''
    labels = labels.split(',')
    n_supp = int(n_supp)

    _search_params['num'] = n_supp
    _search_params['fileType'] = file_type
    _search_params['rights'] = rights
    _search_params['imgType'] = image_type
    _search_params['imgColorType'] = color_type

    fig, axs = plt.subplots(len(labels), n_supp, figsize=(n_supp*4, len(labels)*4))

    with torch.no_grad():
        # query image
        query = preprocess(query).unsqueeze(0).unsqueeze(0).to(device) # (1, 1, 3, H, W)

        supp_x = []
        supp_y = []

        # search support images
        for idx, y in enumerate(labels):
            gis = GoogleImagesSearch(args.api_key, args.cx)
            _search_params['q'] = y
            gis.search(search_params=_search_params, custom_image_name='my_image')
            gis._custom_image_name = 'my_image' # fix: image name sometimes too long

            for j, x in enumerate(gis.results()):
                x.download('./')
                x_im = Image.open(x.path)

                # vis
                axs[idx, j].imshow(x_im)
                axs[idx, j].set_title(f'{y}{j}:{x.url}')
                axs[idx, j].axis('off')

                x_im = preprocess(x_im) # (3, H, W)
                supp_x.append(x_im)
                supp_y.append(idx)

        print('Searching for support images is done.')

        supp_x = torch.stack(supp_x, dim=0).unsqueeze(0).to(device) # (1, n_supp*n_labels, 3, H, W)
        supp_y = torch.tensor(supp_y).long().unsqueeze(0).to(device) # (1, n_supp*n_labels)

        with torch.cuda.amp.autocast(True):
            output = model(supp_x, supp_y, query) # (1, 1, n_labels)

        probs = output.softmax(dim=-1).detach().cpu().numpy()

        return {k: float(v) for k, v in zip(labels, probs[0, 0])}, fig


# DEBUG
##query = Image.open('../labrador-puppy.jpg')
#query = Image.open('/Users/hushell/Documents/Dan_tr.png')
##labels = 'dog, cat'
#labels = 'girl, sussie'
#output = inference(query, labels, n_supp=2)
#print(output)


title = "P>M>F few-shot learning pipeline with Google Image Search (GIS)"
description = "Short description: We take a ViT-small backbone, which is pre-trained with DINO, and meta-trained on Meta-Dataset; for few-shot classification, we use a ProtoNet classifier. The demo can be viewed as zero-shot since the support set is built by searching images from Google. Note that you may need to play with GIS parameters to get good support examples. Besides, GIS is not very stable as search requests may fail for many reasons (e.g., number of requests reaches the limit of the day)."
article = "<p style='text-align: center'><a href='http://arxiv.org/abs/2204.07305' target='_blank'>Arxiv</a></p>"


gr.Interface(fn=inference,
             inputs=[
                 gr.inputs.Image(label="Image to classify", type="pil"),
                 gr.inputs.Textbox(lines=1, label="Class hypotheses:", placeholder="Enter class names separated by ','",),
                 gr.inputs.Slider(minimum=2, maximum=10, step=1, label="GIS: Number of support examples per class"),
                 gr.inputs.Dropdown(['png', 'jpg'], default='png', label='GIS: Image file type'),
                 gr.inputs.Dropdown(['cc_publicdomain', 'cc_attribute', 'cc_sharealike', 'cc_noncommercial', 'cc_nonderived'], default='cc_publicdomain', label='GIS: Copy rights'),
                 gr.inputs.Dropdown(['clipart', 'face', 'lineart', 'stock', 'photo', 'animated', 'imgTypeUndefined'], default='photo', label='GIS: Image type'),
                 gr.inputs.Dropdown(['color', 'gray', 'mono', 'trans', 'imgColorTypeUndefined'], default='color', label='GIS: Image color type'),
             ],
             theme="grass",
             outputs=[
                 gr.outputs.Label(label="Predicted class probabilities"),
                 gr.outputs.Image(type='plot', label="Support examples from Google image search"),
             ],
             title=title,
             description=description,
             article=article,
            ).launch(debug=True)