Table 3 Performance on the ATAS dataset
From: Cellstitch: 3D cellular anisotropic image segmentation via optimal transport
Setting | Method | Precision | Recall | AP | mAP |
|---|---|---|---|---|---|
Anisotropy = 0 | CellStitch | \(0.79 \pm 0.08\) | \(0.77 \pm 0.04\) | \(0.64 \pm 0.07\) | \(0.70 \pm 0.07\) |
Cellpose2D | \(0.61 \pm 0.10\) | \(0.69 \pm 0.04\) | \(0.48 \pm 0.07\) | \(0.54 \pm 0.07\) | |
Cellpose3D\(^*\) | \(\mathbf {0.87 \pm 0.17}\) | \(\mathbf {0.98 \pm 0.01}\) | \(\mathbf {0.85 \pm 0.17}\) | \(\mathbf {0.85 \pm 0.16}\) | |
PlantSeg | \(0.40 \pm 0.07\) | \(0.92 \pm 0.16\) | \(0.39 \pm 0.07\) | \(0.39 \pm 0.07\) | |
Anisotropy = 5 | CellStitch\(^*\) | \(\mathbf {0.81 \pm 0.04}\) | \(\mathbf {0.58 \pm 0.05}\) | \(\mathbf {0.51 \pm 0.05}\) | \(\mathbf {0.62 \pm 0.04}\) |
Cellpose2D | \(0.67 \pm 0.04\) | \(\mathbf {0.61 \pm 0.05}\) | \(0.47 \pm 0.04\) | \(0.56 \pm 0.03\) | |
Cellpose3D | \(0.40 \pm 0.09\) | \(\mathbf {0.59 \pm 0.11}\) | \(0.32 \pm 0.08\) | \(0.32 \pm 0.06\) | |
PlantSeg | \(0.37 \pm 0.08\) | \(0.31 \pm 0.13\) | \(0.20 \pm 0.08\) | \(0.21 \pm 0.06\) | |
Anisotropy = 10 | CellStitch\(^*\) | \(\mathbf {0.75 \pm 0.04}\) | \(0.46 \pm 0.06\) | \(\mathbf {0.40 \pm 0.05}\) | \(\mathbf {0.53 \pm 0.05}\) |
Cellpose2D\(^*\) | \(0.64 \pm 0.04\) | \(\mathbf {0.55 \pm 0.05}\) | \(\mathbf {0.42 \pm 0.05}\) | \(\mathbf {0.52 \pm 0.03}\) | |
Cellpose3D | \(0.30 \pm 0.09\) | \(0.38 \pm 0.11\) | \(0.20 \pm 0.08\) | \(0.26 \pm 0.06\) | |
PlantSeg | \(0.12 \pm 0.06\) | \(0.04 \pm 0.03\) | \(0.03 \pm 0.02\) | \(0.00 \pm 0.00\) |