Introduction
Related work
A review of proposed methods for 3D reconstruction of the colon
An introduction to conditional generative adversarial models
Contribution
Methods
Data generation
Synthetic dataset
Specularities
Generating ground truth data from a phantom
Model
Extension
Method | Mean \(L_1\)-error* | Mean relative \(L_1\)-error** | Mean RMSE* |
---|---|---|---|
Discriminative model |
\(0.234 \pm 0.047\)
|
\( 8.4 \pm 2.1\)
| 0.239 |
pix2pix |
\(0.232 \pm 0.046\)
|
\( 8.2 \pm 2.0\)
| 0.236 |
Extended pix2pix |
\(\mathbf 0.171 \pm 0.034\)
|
\( \mathbf 6.4 \pm 1.7\)
|
\(\mathbf 0.175 \)
|
Training
Experiments
Synthetic data
Real data
Method | Mean \(L_1\)-error* | Mean relative \(L_1\)-error** | Mean RMSE* |
---|---|---|---|
Discriminative model |
\(1.969 \pm 0.040 \)
|
\( 33.9 \pm 0.3\)
|
\(2.207\pm 0.037\)
|
pix2pix |
\( 1.943\pm 0.112\)
|
\( 33.0 \pm 1.8\)
|
\(2.202\pm 0.110\)
|
Extended pix2pix |
\( \mathbf 1.417 \pm 0.128\)
|
\( \mathbf{24.7 } \pm 2.6\)
|
\(\mathbf 1.655 \pm 0.081\)
|