Occlusion Generation

Occlusion generation focuses on creating realistic synthetic data, particularly images with occluded objects, to improve the robustness and accuracy of computer vision models. Current research emphasizes generating diverse and high-quality occlusions, including realistic 3D objects and natural scenarios, often employing deep learning techniques like neural networks to synthesize these occlusions and augment existing datasets. This work is crucial for advancing applications such as pedestrian detection in autonomous driving and face recognition, where handling occlusions is essential for reliable performance. The development of high-quality synthetic occlusion datasets and robust algorithms for handling them is driving significant progress in these fields.

Papers

April 20, 2023

Learning Diffractive Optical Communication Around Arbitrary Opaque Occlusions
Md Sadman Sakib Rahman, Tianyi Gan, Emir Arda Deger, Cagatay Isil, Mona Jarrahi, Aydogan Ozcan
Neural Network Optical Communication Diffractive Processor Occlusion Generation

July 22, 2022

3D Random Occlusion and Multi-Layer Projection for Deep Multi-Camera Pedestrian Localization
Rui Qiu, Ming Xu, Yuyao Yan, Jeremy S. Smith, Xi Yang
Data Augmentation Multi View Pedestrian Detection Multi View Pedestrian Detection Pedestrian Localization Occlusion Generation

May 12, 2022

Delving into High-Quality Synthetic Face Occlusion Segmentation Datasets
Kenny T. R. Voo, Liming Jiang, Chen Change Loy
Unseen Occlusion Occlusion Generation

Occlusion Generation

Papers

Learning Diffractive Optical Communication Around Arbitrary Opaque Occlusions

3D Random Occlusion and Multi-Layer Projection for Deep Multi-Camera Pedestrian Localization

Delving into High-Quality Synthetic Face Occlusion Segmentation Datasets