File size: 2,336 Bytes
83630b2 6a8d727 83630b2 6a8d727 83630b2 6a8d727 83630b2 6a8d727 83630b2 6a8d727 83630b2 6a8d727 83630b2 6a8d727 83630b2 6a8d727 83630b2 6a8d727 83630b2 6a8d727 83630b2 6a8d727 83630b2 |
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 |
```python
!pip install walloc PyWavelets pytorch-wavelets
```
```python
!wget "https://r0k.us/graphics/kodak/kodak/kodim05.png"
```
```python
import os
import torch
import matplotlib.pyplot as plt
import numpy as np
from PIL import Image
from IPython.display import display
from torchvision.transforms import ToPILImage, PILToTensor
from walloc.walloc import Walloc
class Args: pass
```
```python
device = "cpu"
checkpoint = torch.load("v0.6.1.pth",map_location="cpu")
args = checkpoint['args']
codec = Walloc(
channels = args.channels,
J = args.J,
N = args.N,
latent_dim = args.latent_dim,
latent_bits = 5
)
codec.load_state_dict(checkpoint['model_state_dict'])
codec = codec.to(device)
```
```python
img = Image.open("kodim05.png")
img
```
```python
with torch.no_grad():
codec.eval()
x = PILToTensor()(img).to(torch.float)
x = (x/255 - 0.5).unsqueeze(0).to(device)
x_hat, _, _ = codec(x)
ToPILImage()(x_hat[0]+0.5)
```
```python
with torch.no_grad():
codec.eval()
X = codec.wavelet_analysis(x,J=codec.J)
Y = codec.encoder(X)
X_hat = codec.decoder(Y)
x_hat = codec.wavelet_synthesis(X_hat,J=codec.J)
print(f"dimensionality reduction: {x.numel()/Y.numel()}x")
```
dimensionality reduction: 12.0x
```python
Y.unique()
```
tensor([-15., -14., -13., -12., -11., -10., -9., -8., -7., -6., -5., -4.,
-3., -2., -1., 0., 1., 2., 3., 4., 5., 6., 7., 8.,
9., 10., 11., 12., 13., 14., 15.])
```python
plt.hist(Y.flatten().numpy(),range=(-17.5,17.5),bins=35);
```
```python
grid_size = 4
n_channels, H, W = Y[0].shape
combined_image = Image.new('L', (W * grid_size, H * grid_size))
size_bytes = 0
for i, channel in enumerate(Y[0]):
channel = (channel+16).to(torch.uint8)
row = i // grid_size
col = i % grid_size
channel = ToPILImage()(channel)
combined_image.paste(channel, (col * W, row * H))
combined_image
```
```python
combined_image.save('tmp.png')
print("compression_ratio: ", x.numel()/os.path.getsize("tmp.png"))
```
compression_ratio: 20.75383532723434
|