File size: 2,385 Bytes
d4ab5ac
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8fd2935
 
d4ab5ac
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8fd2935
d4ab5ac
d85fbeb
8fd2935
 
 
 
d4ab5ac
 
56ec0e7
d85fbeb
 
8fd2935
d4ab5ac
 
 
d85fbeb
d4ab5ac
 
 
 
8fd2935
 
d85fbeb
d4ab5ac
 
 
8fd2935
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
import sys
sys.path.insert(0, './code')

from datamodules.transformations import UnNest
from models.interpretation import ImageInterpretationNet
from transformers import ViTFeatureExtractor, ViTForImageClassification
from utils.plot import smoothen, draw_mask_on_image, draw_heatmap_on_image

import gradio as gr
import numpy as np
import torch

# Load Vision Transformer
hf_model = "tanlq/vit-base-patch16-224-in21k-finetuned-cifar10"
vit = ViTForImageClassification.from_pretrained(hf_model)
vit.eval()

# Load Feature Extractor
feature_extractor = ViTFeatureExtractor.from_pretrained(hf_model, return_tensors="pt")
feature_extractor = UnNest(feature_extractor)

# Load Vision DiffMask
diffmask = ImageInterpretationNet.load_from_checkpoint('checkpoints/diffmask.ckpt')
diffmask.set_vision_transformer(vit)
diffmask_imagenet = ImageInterpretationNet.load_from_checkpoint('checkpoints/diffmask_imagenet.ckpt')
diffmask_imagenet.set_vision_transformer(vit)

# Define mask plotting functions
def draw_mask(image, mask):
    return draw_mask_on_image(image, smoothen(mask))\
        .permute(1, 2, 0)\
        .clip(0, 1)\
        .numpy()


def draw_heatmap(image, mask):
    return draw_heatmap_on_image(image, smoothen(mask))\
        .permute(1, 2, 0)\
        .clip(0, 1)\
        .numpy()


# Define callable method for the demo
def get_mask(image, model_name: str):
    if image is None:
        return None, None
    if model_name == 'DiffMask-CiFAR-10':
        diffmask_model = diffmask
    elif model_name == 'DiffMask-ImageNet':
        diffmask_model = diffmask_imagenet
    image = torch.from_numpy(image).permute(2, 0, 1).float() / 255
    dm_image = feature_extractor(image).unsqueeze(0)
    dm_out = diffmask_model.get_mask(dm_image)
    mask = dm_out["mask"][0].detach()
    pred = dm_out["pred_class"][0].detach()
    pred = diffmask_model.model.config.id2label[pred.item()]

    masked_img = draw_mask(image, mask)
    heatmap = draw_heatmap(image, mask)
    return np.hstack((masked_img, heatmap)), pred

# Launch demo interface
gr.Interface(
    get_mask,
    inputs=[gr.inputs.Image(label="Input", shape=(224, 224), source="upload", type="numpy"),
    gr.inputs.Dropdown(["DiffMask-CiFAR-10", "DiffMask-ImageNet"])],
    outputs=[gr.outputs.Image(label="Output"), gr.outputs.Label(label="Prediction")],
    title="Vision DiffMask Demo",
    live=True,
).launch()