MLXID/DeepLearning
0
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Rename single photon files

Browse Source
This commit is contained in:
xie_x1
2026-03-18 10:50:08 +01:00
parent 341940ca13
commit 96e78f8ab5
2 changed files with 26 additions and 24 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
Inference_SinglePhoton.py → Infer_1Photon.py
View File

@@ -38,7 +38,7 @@ task = 'SiemenStarLowerLeft'
config = configs[task]
BinningFactor = 10
numberOfAugOps = 1
numberOfAugOps = 8
flag_normalize = False
Roi = config['roi']
X_st, X_ed, Y_st, Y_ed = Roi
@@ -76,7 +76,7 @@ def apply_inverse_transforms(predictions: torch.Tensor, numberOfAugOps: int) ->
if __name__ == "__main__":
model = models.get_model_class(config['modelVersion'])().cuda()
modelName = f'singlePhoton{config["modelVersion"]}_{config["energy"]}keV_Noise{config["noise"]}keV_aug1'
modelName = f'singlePhoton{config["modelVersion"]}_{config["energy"]}keV_Noise{config["noise"]}keV_E500_aug1'
if flag_normalize:
modelName += '_normalized'
model.load_state_dict(torch.load(f'/home/xie_x1/MLXID/DeepLearning/Models/{modelName}.pth', weights_only=True))

46
Train_SinglePhoton.py → Train_1Photon.py
View File

@@ -23,7 +23,7 @@ TestLoss = -1
model = models.get_model_class(modelVersion)().cuda()
# summary(model, input_size=(128, 1, 3, 3))
LearningRate = 1e-3
Noise = 0.13 # in keV
Noise = 0.23 # in keV
NoiseThreshold = 0 * Noise # in keV, set values below this threshold to zero
numberOfAugOps = 1 # 1 (no augmentation) or (1,8] (with augmentation)
flag_normalize = False
@@ -141,27 +141,8 @@ testLoader = torch.utils.data.DataLoader(
optimizer = torch.optim.Adam(model.parameters(), lr=LearningRate)
scheduler = optim.lr_scheduler.ReduceLROnPlateau(optimizer, 'min', factor=0.7, patience = 3)
if __name__ == "__main__":
for epoch in tqdm(range(1, 1001)):
train(model, trainLoader, optimizer)
test(model, valLoader)
scheduler.step(ValLosses[-1])
print(f"Learning Rate: {optimizer.param_groups[0]['lr']}")
if epoch in [20, 30, 50, 100, 200, 300, 500, 750, 1000]:
modelName = f'singlePhoton{modelVersion}_{Energy}_Noise{Noise}keV_E{epoch}_aug{numberOfAugOps}'
if flag_normalize == True:
modelName += '_normalized'
torch.save(model.state_dict(), f'Models/{modelName}.pth')
print(f"Saved model checkpoint: {modelName}.pth")
test(model, testLoader)
modelName = f'singlePhoton{modelVersion}_{Energy}_Noise{Noise}keV_E{epoch}_aug{numberOfAugOps}'
if flag_normalize == True:
modelName += '_normalized'
torch.save(model.state_dict(), f'Models/{modelName}.pth')
print(f"Saved final model checkpoint: {modelName}.pth")
def plot_loss_curve(TrainLosses, ValLosses, TestLoss, modelVersion):
def plot_loss_curve(TrainLosses, ValLosses, modelVersion, TestLoss=0):
import matplotlib.pyplot as plt
plt.figure(figsize=(8,6))
plt.plot(TrainLosses, label='Train Loss', color='blue')
@@ -177,4 +158,25 @@ def plot_loss_curve(TrainLosses, ValLosses, TestLoss, modelVersion):
if flag_normalize:
plotName += '_normalized'
plt.savefig(f'Results/{plotName}.png')
plot_loss_curve(TrainLosses, ValLosses, TestLoss, modelVersion=modelVersion)
if __name__ == "__main__":
for epoch in tqdm(range(1, 151)):
train(model, trainLoader, optimizer)
test(model, valLoader)
scheduler.step(ValLosses[-1])
print(f"Learning Rate: {optimizer.param_groups[0]['lr']}")
if epoch in [20, 30, 50, 100, 150]:
modelName = f'singlePhoton{modelVersion}_{Energy}_Noise{Noise}keV_E{epoch}_aug{numberOfAugOps}'
if flag_normalize == True:
modelName += '_normalized'
torch.save(model.state_dict(), f'Models/{modelName}.pth')
print(f"Saved model checkpoint: {modelName}.pth")
plot_loss_curve(TrainLosses, ValLosses, modelVersion=modelVersion)
test(model, testLoader)
modelName = f'singlePhoton{modelVersion}_{Energy}_Noise{Noise}keV_E{epoch}_aug{numberOfAugOps}'
if flag_normalize == True:
modelName += '_normalized'
torch.save(model.state_dict(), f'Models/{modelName}.pth')
print(f"Saved final model checkpoint: {modelName}.pth")
plot_loss_curve(TrainLosses, ValLosses, modelVersion=modelVersion, TestLoss=TestLoss)