Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,61 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
datasets:
|
| 3 |
+
- eurecom-ds/shapes3d
|
| 4 |
+
library_name: diffusers
|
| 5 |
+
pipeline_tag: Conditional Image Generation
|
| 6 |
+
---
|
| 7 |
+
|
| 8 |
+
```python
|
| 9 |
+
# !pip install diffusers
|
| 10 |
+
from diffusers import DiffusionPipeline
|
| 11 |
+
import torch
|
| 12 |
+
|
| 13 |
+
from PIL import Image
|
| 14 |
+
|
| 15 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
| 16 |
+
model_id = "eurecom-ds/scoresdeve-conditional-ema-shapes3d-64"
|
| 17 |
+
|
| 18 |
+
# load model and scheduler
|
| 19 |
+
pipe = DiffusionPipeline.from_pretrained(model_id, trust_remote_code=True)
|
| 20 |
+
pipe.to(device)
|
| 21 |
+
|
| 22 |
+
|
| 23 |
+
# run pipeline in inference (sample random noise and denoise)
|
| 24 |
+
generator = torch.Generator(device=device).manual_seed(46)
|
| 25 |
+
class_labels = torch.tensor([[0, 0, 0, 0, 1, 0], # condition on shape cube
|
| 26 |
+
[0, 0, 0, 0, 2, 0], # condition on shape cylinder
|
| 27 |
+
[0, 0, 0, 0, 3, 0], # condition on shape sphere
|
| 28 |
+
[0, 0, 0, 0, 4, 0], # condition on shape capsule
|
| 29 |
+
[0, 0, 0, 0, 0, 0], # unconditional
|
| 30 |
+
[1, 1, 1, 1, 1, 1], # condition on red floor, object red, orientation right, small scale, shape cube, wall red
|
| 31 |
+
[0, 0, 0, 0, 0, 0], # unconditional
|
| 32 |
+
[0, 0, 0, 0, 0, 0], # uncondtional
|
| 33 |
+
[0, 0, 0, 0, 1, 0], # condition on shape cube
|
| 34 |
+
[0, 0, 0, 0, 1, 0], # condition on shape cube
|
| 35 |
+
[0, 0, 0, 0, 1, 0], # condition on shape cube
|
| 36 |
+
[0, 0, 0, 0, 1, 0], # condition on shape cube
|
| 37 |
+
[0, 0, 0, 0, 1, 0], # condition on shape cube
|
| 38 |
+
[0, 0, 0, 0, 1, 0], # condition on shape cube
|
| 39 |
+
[0, 0, 0, 0, 1, 0], # condition on shape cube
|
| 40 |
+
[0, 0, 0, 0, 1, 0] # condition on shape cube
|
| 41 |
+
]).to(device=pipe.device)
|
| 42 |
+
image = pipe(
|
| 43 |
+
generator=generator,
|
| 44 |
+
batch_size=16,
|
| 45 |
+
class_labels=class_labels,
|
| 46 |
+
num_inference_steps=1000
|
| 47 |
+
).images
|
| 48 |
+
width, height = image[0].size
|
| 49 |
+
|
| 50 |
+
# Create a new image with enough space for 2 rows x 8 columns
|
| 51 |
+
grid = Image.new('RGB', (width * 8, height * 2))
|
| 52 |
+
|
| 53 |
+
for index, img in enumerate(image):
|
| 54 |
+
x = index % 8 * width # Column index (0-7) times width of one image
|
| 55 |
+
y = index // 8 * height # Row index (0-1) times height of one image
|
| 56 |
+
grid.paste(img, (x, y))
|
| 57 |
+
|
| 58 |
+
# Save the final grid image
|
| 59 |
+
grid.save("sde_ve_conditional_generated_grid.png")
|
| 60 |
+
```
|
| 61 |
+
|