-
Notifications
You must be signed in to change notification settings - Fork 0
/
tsne_interactive.py
78 lines (60 loc) · 2.5 KB
/
tsne_interactive.py
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
from sklearn.manifold import TSNE
import numpy as np
import torch
import matplotlib.pyplot as plt
from matplotlib.offsetbox import OffsetImage, AnnotationBbox
from lib.models import load_backbone
from lib.datasets import get_dataloader
from config import get_args
def main(args):
backbone = load_backbone(model=args.backbone_model)
test_dataloader = get_dataloader(batch_size=64, shuffle=False, num_workers=4, mode="test", split=args.split, model=args.backbone_model)
tsne = TSNE(n_components=2, random_state=42)
test_images = []
test_embbedings = []
test_labels = []
with torch.no_grad():
for imgs, labels in test_dataloader:
imgs = imgs.cuda()
labels = labels.cuda()
embeddings = backbone(imgs)
imgs_np = ((imgs.cpu().numpy()*0.5+0.5)*255).transpose(0, 2, 3, 1)
imgs_np = imgs_np.astype(np.uint8)
test_images.append(imgs_np)
test_embbedings.append(embeddings.cpu().numpy())
test_labels.append(labels.cpu().numpy())
test_images = np.concatenate(test_images)
test_embbedings = np.concatenate(test_embbedings)
test_labels = np.concatenate(test_labels)
X_2d = tsne.fit_transform(test_embbedings)
cmap = plt.cm.get_cmap('tab20', 20)
fig, (ax_scatter, ax_image) = plt.subplots(1, 2, figsize=(12, 6))
scatter = ax_scatter.scatter(X_2d[:, 0], X_2d[:, 1], s=10, c=test_labels, cmap=cmap)
ax_scatter.set_xlabel('t-SNE feature 1')
ax_scatter.set_ylabel('t-SNE feature 2')
ax_scatter.set_title(f't-SNE visualization {args.backbone_model}')
ax_image.imshow(np.ones((224, 224, 3)), aspect="auto")
ax_image.axis('off')
circle = ax_scatter.scatter(0, 0, s=10, c="k")
circle.set_visible(False)
def hover(event):
is_contained, annotation_index = scatter.contains(event)
if is_contained:
data_index = annotation_index['ind'][0]
ax_image.clear()
ax_image.imshow(test_images[data_index], aspect='auto')
ax_image.axis('off')
x, y = X_2d[data_index]
circle.set_offsets([x, y])
circle.set_visible(True)
fig.canvas.draw_idle()
else:
ax_image.imshow(np.ones((224, 224, 3)), aspect='auto')
ax_image.axis('off')
circle.set_visible(False)
fig.canvas.draw_idle()
fig.canvas.mpl_connect("motion_notify_event", hover)
plt.show()
if __name__ == "__main__":
args = get_args()
main(args)