trung-vt/ddct

⬚ MODEL_NAME

Short, one-sentence overview of what the model does and why it matters.

---

Table of Contents

1. Model Details
2. Installation
3. Quick Start
4. Training & Evaluation
5. Results
6. Limitations & Ethical Considerations
7. Citation
8. Contact

---

🔖 Pre-trained checkpoints

Alias	Direct link	File in repo
LPD fan 360 angles	epoch 099	results/logs/mayo_ct_fan_beam_360_angles_full/learned/LPD/LPD_lr0.001/version_0/checkpoints/epoch=099.ckpt
FBPConvNet fan 360 angles	epoch 024	results/logs/mayo_ct_fan_beam_360_angles_full/learned/FBPConvNet/csv/FBPConvNet/version_0/checkpoints/epoch%3D024.ckpt
AR fan 360 angles	epoch 099	results/logs/mayo_ct_fan_beam_360_angles_full/AR/csv/AR_reg_lr0.0001_lambda_gp10/version_0/checkpoints/epoch=099.ckpt
UAR fan 360 angles (both λ)	epoch 049	results/logs/mayo_ct_fan_beam_360_angles_full/UAR/csv/UAR_reg_lr0.0001_gen_lr0.0001_lambda_gp10.0_lambda_vp0.1/version_0/checkpoints/epoch=049.ckpt
AR parallel 200 angles	epoch 029	results/logs/mayo_ct_parallel_beam_200_angles_full/learned/AR/AR_reg_lr0.0001_lambda_gp1/version_1_20250601_good/checkpoints/epoch_epoch=029.ckpt
UAR parallel 200 angles (learned both $\lambda_\text{primal}$ and $\lambda_\text{dual}$)	epoch 059	results/logs/mayo_ct_parallel_beam_200_angles_full/learned/UAR/UAR_reg_lr0.0001_gen_lr0.0001_lambda_gp10.0_lambda_vp0.1/version_8_both_learned_lambdas_primal_dual_good_20250611/checkpoints/epoch_epoch=059.ckpt
UAR parallel 200 angles (single λ)	epoch 059	results/logs/mayo_ct_parallel_beam_200_angles_full/learned/UAR/UAR_reg_lr0.0001_gen_lr0.0001_lambda_gp10.0_lambda_vp0.1/version_3_single_fixed_lambda_primal_good_20250610/checkpoints/epoch_epoch=059.ckpt

Python download snippets

from huggingface_hub import hf_hub_download

# LPD fan 360°
path = hf_hub_download(
    repo_id="trung-vt/ddct",
    filename="epoch=099.ckpt",
    subfolder=(
        "results/logs/mayo_ct_fan_beam_360_angles_full/learned/"
        "LPD/LPD_lr0.001/version_0/checkpoints"
    ),
)
print(path)

---

Model Details

Item	Description
Architecture	⬚ e.g. Learned Primal-Dual (LPD), 12 blocks
Input	⬚ Sinogram (60 fan-beam angles, 512 × 512)
Output	⬚ Reconstructed CT slice
Params	⬚ # of parameters
Training data	Mayo Clinic Low-Dose CT fan-beam subset
Checkpoint(s)	epoch=4-step=9999.ckpt, best SSIM, etc.

---

Installation

pip install huggingface-hub pytorch-lightning                # core deps
pip install ⬚ any-extra-package-your-model-needs