| START |
| INPUT: NS, annotation (orientation) |
OUTPUT: Localized RoI, CMskDenseNet-77 |
NS : Total image samples containing. |
annotation (orientation): Mask coordinates of the glaucoma regions in the retinal image |
Localized RoI : Region placement |
CMskDenseNet-77- : Custom Mask-RCNN network with DenseNe-77 key points |
SampleResolution ← [x y] |
// Computing Mask |
µ← AnchorsComputation (NS, annotation) |
// Customized MaskRCNN model |
CMskDenseNet-77← DesignCustomDenseNet-77MaskRCNN (SampleResolution, µ) |
[ Sr, St] ← database division into train and test section |
| // Glaucoma Region recognition from Training part |
For each sample f in ⟶Sr |
Compute DenseNet-77 keypoints ⟶ns |
End For |
Training CMskDenseNet-77over ns, and compute training time t_dense |
∂_dense ← PreRegionLoc(ns) |
Ap_dense ← Evaluate_AP (DenseNet-77, ∂_dense) |
For each sample F in ⟶ St |
| (a) compute features by employing trained model ¥⟶βI |
(b) [Mask, objectness_score, classLabel] ←Predict (βI) |
(c) Output sample along with Mask, class |
(d) ∂← [∂ Mask] |
End For |
Ap_¥← Evaluate framework ¥ using ∂ |
FINISH. |