File size: 9,998 Bytes
dcd2bd2 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 | import argparse
import os
import torch
from PIL import Image
from torchvision import transforms
from train_network import (
DEFAULT_GAMMAS,
DEFAULT_SIGMA_LEVELS,
DEFAULT_TAU_INITS,
_collect_image_paths,
_SCRIPT_DIR,
calculate_psnr,
gamma_tag,
sigma_int_to_float,
sigma_tag,
tau_tag,
)
from train_network import UnrolledNetwork as MLPUnrolledNetwork
from train_network_rbf import UnrolledNetwork as RBFUnrolledNetwork
from train_tnrd_baseline import TNRDBaselineNetwork
DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
TESTSETS = ("Set12", "BSD68")
def _testset_root(name):
return os.path.join(
_SCRIPT_DIR,
"datasets",
"Test_Datasets",
"FFDNet-master",
"testsets",
name,
)
def _autocast_context():
return (
torch.amp.autocast("cuda")
if DEVICE.type == "cuda"
else torch.autocast("cpu", enabled=False)
)
def _build_model(model_type, stages, use_wave, damping_gamma, tau_init):
if model_type == "mlp":
return MLPUnrolledNetwork(stages, use_wave, damping_gamma=damping_gamma, tau_init=tau_init).to(DEVICE)
if model_type == "rbf":
return RBFUnrolledNetwork(stages, use_wave, damping_gamma=damping_gamma, tau_init=tau_init).to(DEVICE)
if model_type == "tnrd":
return TNRDBaselineNetwork(stages, tau_init=tau_init).to(DEVICE)
raise ValueError(f"Unknown model type: {model_type}")
def _checkpoint_specs(stages, sigmas, gammas, tau_inits, include_finetuned):
specs = []
for sigma in sigmas:
sigma_name = sigma_tag(sigma)
for tau_init in tau_inits:
tau_name = tau_tag(tau_init)
specs.append(
{
"label": f"TNRD baseline sigma={sigma} tau={tau_init}",
"model_type": "tnrd",
"use_wave": False,
"damping_gamma": 1.0,
"tau_init": tau_init,
"path": f"tnrd_baseline_{stages}stages_{sigma_name}_{tau_name}.pth",
}
)
if include_finetuned:
specs.append(
{
"label": f"Finetuned TNRD baseline sigma={sigma} tau={tau_init}",
"model_type": "tnrd",
"use_wave": False,
"damping_gamma": 1.0,
"tau_init": tau_init,
"path": f"finetuned_tnrd_baseline_{stages}stages_{sigma_name}_{tau_name}.pth",
}
)
for damping_gamma in gammas:
gamma_name = gamma_tag(damping_gamma)
for tau_init in tau_inits:
tau_name = tau_tag(tau_init)
specs.extend(
[
{
"label": f"MLP Telegraph sigma={sigma} gamma={damping_gamma} tau={tau_init}",
"model_type": "mlp",
"use_wave": True,
"damping_gamma": damping_gamma,
"tau_init": tau_init,
"path": f"model_{stages}stages_waveTrue_{sigma_name}_{gamma_name}_{tau_name}.pth",
},
{
"label": f"MLP No-wave sigma={sigma} gamma={damping_gamma} tau={tau_init}",
"model_type": "mlp",
"use_wave": False,
"damping_gamma": damping_gamma,
"tau_init": tau_init,
"path": f"model_{stages}stages_waveFalse_{sigma_name}_{gamma_name}_{tau_name}.pth",
},
{
"label": f"RBF Telegraph sigma={sigma} gamma={damping_gamma} tau={tau_init}",
"model_type": "rbf",
"use_wave": True,
"damping_gamma": damping_gamma,
"tau_init": tau_init,
"path": f"rbf_model_{stages}stages_waveTrue_{sigma_name}_{gamma_name}_{tau_name}.pth",
},
{
"label": f"RBF No-wave sigma={sigma} gamma={damping_gamma} tau={tau_init}",
"model_type": "rbf",
"use_wave": False,
"damping_gamma": damping_gamma,
"tau_init": tau_init,
"path": f"rbf_model_{stages}stages_waveFalse_{sigma_name}_{gamma_name}_{tau_name}.pth",
},
]
)
if include_finetuned:
specs.extend(
[
{
"label": f"Finetuned MLP Telegraph sigma={sigma} gamma={damping_gamma} tau={tau_init}",
"model_type": "mlp",
"use_wave": True,
"damping_gamma": damping_gamma,
"tau_init": tau_init,
"path": f"finetuned_{stages}stages_waveTrue_{sigma_name}_{gamma_name}_{tau_name}.pth",
},
{
"label": f"Finetuned MLP No-wave sigma={sigma} gamma={damping_gamma} tau={tau_init}",
"model_type": "mlp",
"use_wave": False,
"damping_gamma": damping_gamma,
"tau_init": tau_init,
"path": f"finetuned_{stages}stages_waveFalse_{sigma_name}_{gamma_name}_{tau_name}.pth",
},
{
"label": f"Finetuned RBF Telegraph sigma={sigma} gamma={damping_gamma} tau={tau_init}",
"model_type": "rbf",
"use_wave": True,
"damping_gamma": damping_gamma,
"tau_init": tau_init,
"path": f"finetuned_rbf_model_{stages}stages_waveTrue_{sigma_name}_{gamma_name}_{tau_name}.pth",
},
{
"label": f"Finetuned RBF No-wave sigma={sigma} gamma={damping_gamma} tau={tau_init}",
"model_type": "rbf",
"use_wave": False,
"damping_gamma": damping_gamma,
"tau_init": tau_init,
"path": f"finetuned_rbf_model_{stages}stages_waveFalse_{sigma_name}_{gamma_name}_{tau_name}.pth",
},
]
)
return specs
def evaluate_checkpoint(spec, dataset_name, sigma, stages):
model = _build_model(
spec["model_type"],
stages,
spec["use_wave"],
spec["damping_gamma"],
spec["tau_init"],
)
state = torch.load(spec["path"], map_location=DEVICE)
model.load_state_dict(state)
model.eval()
test_root = _testset_root(dataset_name)
test_paths = _collect_image_paths(test_root)
if not test_paths:
raise FileNotFoundError(
f"No test images found in {os.path.abspath(test_root)} for {dataset_name}."
)
sigma_float = sigma_int_to_float(sigma)
test_transform = transforms.Compose([transforms.Grayscale(), transforms.ToTensor()])
torch.manual_seed(42)
total_psnr = 0.0
with torch.no_grad():
for path in test_paths:
clean = test_transform(Image.open(path)).unsqueeze(0).to(DEVICE)
noisy = torch.clamp(clean + torch.randn_like(clean) * sigma_float, 0.0, 1.0)
with _autocast_context():
output = model(noisy)
total_psnr += calculate_psnr(clean, output)
return total_psnr / len(test_paths)
def main(args):
specs = _checkpoint_specs(
args.stages,
[int(s) for s in args.sigmas],
[float(g) for g in args.gammas],
[float(t) for t in args.tau_inits],
args.include_finetuned,
)
print(f"[*] Evaluating checkpoints on {', '.join(TESTSETS)}")
print(f"[*] Device: {DEVICE}")
print("-" * 90)
print(f"{'Model':<38} {'Dataset':<8} {'PSNR':>8} Checkpoint")
print("-" * 90)
for spec in specs:
if not os.path.exists(spec["path"]):
print(f"{spec['label']:<38} {'-':<8} {'[missing]':>8} {spec['path']}")
continue
sigma_value = int(spec["label"].split("sigma=")[-1])
for dataset_name in TESTSETS:
psnr = evaluate_checkpoint(spec, dataset_name, sigma_value, args.stages)
print(f"{spec['label']:<38} {dataset_name:<8} {psnr:>8.2f} {spec['path']}")
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--stages", type=int, default=5)
parser.add_argument(
"--sigmas",
type=int,
nargs="+",
default=list(DEFAULT_SIGMA_LEVELS),
help="Noise levels to evaluate, specified in 0-255 units.",
)
parser.add_argument(
"--gammas",
type=float,
nargs="+",
default=list(DEFAULT_GAMMAS),
help="Fixed damping gamma values to evaluate for MLP/RBF models.",
)
parser.add_argument(
"--tau_inits",
type=float,
nargs="+",
default=list(DEFAULT_TAU_INITS),
help="Initial tau values to evaluate.",
)
parser.add_argument(
"--include_finetuned",
action="store_true",
help="Also evaluate finetuned checkpoints.",
)
args = parser.parse_args()
main(args)
|