File size: 4,603 Bytes
01fdb75 | 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 | import sys
sys.path.insert(0, "/data/sichengli/Code/PixelGen")
import torch
import numpy as np
from PIL import Image, ImageDraw, ImageFont
import torchvision.transforms as transforms
import torchvision.transforms.functional as TF
import os, random
from src.models.transformer.JiT_medical import JiTMedical
from src.diffusion.flow_matching.sampling_medical import EulerSamplerMedical
from src.diffusion.flow_matching.scheduling import LinearScheduler
from src.diffusion.base.guidance import simple_guidance_fn
device = torch.device("cuda:0")
# Load model
ckpt_path = "/data/sichengli/Code/PixelGen/medical_workdirs/exp_PixelGen_Medical_B16/epoch=236-step=100000.ckpt"
ckpt = torch.load(ckpt_path, map_location="cpu", weights_only=False)
state_dict = ckpt["state_dict"]
model = JiTMedical(
input_size=256, patch_size=16, in_channels=3,
hidden_size=768, depth=12, num_heads=12, mlp_ratio=4.0,
attn_drop=0.0, proj_drop=0.1, num_classes=1,
use_bottleneck=True, bottleneck_dim=128,
in_context_len=32, in_context_start=4, mask_in_channels=1
)
ema_state = {}
for k, v in state_dict.items():
if k.startswith("ema_denoiser."):
new_k = k.replace("ema_denoiser.", "").replace("_orig_mod.", "")
ema_state[new_k] = v
model.load_state_dict(ema_state, strict=False)
model = model.to(device).eval().to(torch.float32)
sampler = EulerSamplerMedical(
num_steps=50, guidance=2.0, timeshift=1.0,
guidance_interval_min=0.1, guidance_interval_max=0.9,
scheduler=LinearScheduler(), w_scheduler=LinearScheduler(),
guidance_fn=simple_guidance_fn,
).to(device)
# Load 8 samples
data_root = "/data2/sichengli/Data/test/Segmentation/OCTA500"
img_dir = os.path.join(data_root, "images")
mask_dir = os.path.join(data_root, "masks")
all_files = sorted([f for f in os.listdir(img_dir) if f.endswith(".png") and not f.startswith("thumb")])
random.seed(456)
selected = random.sample(all_files, 8)
images_list, masks_list = [], []
for fname in selected:
img = Image.open(os.path.join(img_dir, fname)).convert("L")
img = TF.resize(img, (256, 256))
images_list.append(TF.to_tensor(img).repeat(3, 1, 1))
mask = Image.open(os.path.join(mask_dir, fname)).convert("L")
mask = TF.resize(mask, (256, 256), interpolation=transforms.InterpolationMode.NEAREST)
masks_list.append(TF.to_tensor(mask))
real_images = torch.stack(images_list)
masks_tensor = torch.stack(masks_list)
# Generate
print("Generating...")
with torch.no_grad():
noise = torch.randn(8, 3, 256, 256, device=device)
x_trajs, _ = sampler._impl_sampling(model, noise, None, None, mask=masks_tensor.to(device))
generated = x_trajs[-1].clamp(-1, 1) * 0.5 + 0.5
# Create comparison: 8 rows x 3 columns (Mask | Generated | Real)
h, w = 256, 256
pad = 6
label_h = 40
n_rows = 8
n_cols = 3
canvas_w = n_cols * w + (n_cols + 1) * pad
canvas_h = n_rows * h + (n_rows + 1) * pad + label_h
canvas = np.ones((canvas_h, canvas_w, 3), dtype=np.uint8) * 30
# Add column labels
labels = ["Mask", "Generated", "Real"]
try:
font = ImageFont.truetype("/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf", 24)
except Exception:
font = ImageFont.load_default()
pil_canvas = Image.fromarray(canvas)
draw = ImageDraw.Draw(pil_canvas)
for col, label in enumerate(labels):
x_pos = pad + col * (w + pad) + w // 2
bbox = draw.textbbox((0, 0), label, font=font)
text_w = bbox[2] - bbox[0]
draw.text((x_pos - text_w // 2, 8), label, fill=(255, 255, 255), font=font)
canvas = np.array(pil_canvas)
# Colormap for mask classes
color_map = {
0: (0, 0, 0),
50: (255, 80, 80),
100: (80, 255, 80),
150: (80, 80, 255),
200: (255, 255, 80),
250: (255, 80, 255),
}
for row in range(n_rows):
y = label_h + pad + row * (h + pad)
# Mask - colorized
m = masks_tensor[row, 0].numpy()
m_uint8 = (m * 255).astype(np.uint8)
m_colored = np.zeros((h, w, 3), dtype=np.uint8)
for val, color in color_map.items():
mask_region = np.abs(m_uint8.astype(int) - val) < 13
m_colored[mask_region] = color
canvas[y:y+h, pad:pad+w] = m_colored
# Generated
g = generated[row].cpu().permute(1, 2, 0).numpy()
g = (g * 255).clip(0, 255).astype(np.uint8)
canvas[y:y+h, 2*pad+w:2*pad+2*w] = g
# Real
r = real_images[row].permute(1, 2, 0).numpy()
r = (r * 255).clip(0, 255).astype(np.uint8)
canvas[y:y+h, 3*pad+2*w:3*pad+3*w] = r
out = "/data/sichengli/Code/PixelGen/medical_workdirs/exp_PixelGen_Medical_B16/val_samples/mask_control_final.png"
Image.fromarray(canvas).save(out)
print("Saved:", out, canvas.shape)
|