Deep Object Occlusion Relationship Detection Based on Associative Embedding Clustering DOI Creative Commons

Peiyong Gong,

Kai Zheng, Ting Liu

и другие.

Technologies, Год журнала: 2025, Номер 13(4), С. 143 - 143

Опубликована: Апрель 4, 2025

Visual relationship detection is crucial for understanding scenes depicted in images when aiming to detect objects within the image and recognize visual relationships between each pair of objects. Nevertheless, profound occlusion, as a typical existing constituting pivotal semantic feature, has regrettably been subjected insufficient scrutiny. To address this issue, we propose pioneering approach termed DOORD-AEC, which specifically designed detecting occlusion spatial among targets. DOORD-AEC introduces associative embedding clustering supervise convolutional neural network with two branches, enabling it take an input produce triplet set representing relationships. The learns simultaneously identify all targets occlusions that make up group them together using clustering. Additionally, contribute KORD dataset, novel challenging dataset We demonstrate effectiveness our method dataset.

Язык: Английский

Deep Object Occlusion Relationship Detection Based on Associative Embedding Clustering DOI Creative Commons

Peiyong Gong,

Kai Zheng, Ting Liu

и другие.

Technologies, Год журнала: 2025, Номер 13(4), С. 143 - 143

Опубликована: Апрель 4, 2025

Visual relationship detection is crucial for understanding scenes depicted in images when aiming to detect objects within the image and recognize visual relationships between each pair of objects. Nevertheless, profound occlusion, as a typical existing constituting pivotal semantic feature, has regrettably been subjected insufficient scrutiny. To address this issue, we propose pioneering approach termed DOORD-AEC, which specifically designed detecting occlusion spatial among targets. DOORD-AEC introduces associative embedding clustering supervise convolutional neural network with two branches, enabling it take an input produce triplet set representing relationships. The learns simultaneously identify all targets occlusions that make up group them together using clustering. Additionally, contribute KORD dataset, novel challenging dataset We demonstrate effectiveness our method dataset.

Язык: Английский

Процитировано

0