U-Net with Hierarchical Bottleneck Attention for Landmark Detection in Fundus Images of the Degenerated Retina

Please cite as:

S Tang, Z Qi, J Granley, M Beyeler (2021). U-Net with Hierarchical Bottleneck Attention for Landmark Detection in Fundus Images of the Degenerated Retina. MICCAI OMIA8 Workshop, online.

The preprint can be found on arXiv and the published paper here.

Check out Papers with Code to see the Global Rankings. As of January 26, 2023, HBA-U-Net is still listed as the state of the art (SOTA) for several popular datasets of retinal degeneration:

• #1 ADAM: Fovea detection
• #1 ADAM: Optic disc segmentation
• #1 IDRiD: Fovea detection
• #1 IDRiD: Optic disc detection
• #1 REFUGE: Fovea detection
• #2 REFUGE: Optic disc segmentation

Requirements

• Python 3
• Keras 2.4.3
• TensorFlow 2.5.0
• Scikit-Learn 0.22
• Skimage 0.16.2
• cv2 4.1.2
• PIL 7.1.2
• Pandas 1.1.5

How to acquire the private dataset listed in the paper

• ADAM
• REFUGE
• IDRiD