Human-centered deep compositional model for handling occlusions

Koporec, Gregor (avtor), Perš, Janez (avtor)

Despite their powerful discriminative abilities, Convolutional Neural Networks (CNNs) lack the properties of generative models. This leads to a decreased performance in environments where objects are poorly visible. Solving such a problem by adding more training samples can quickly lead to a combinatorial explosion, therefore the underlying architecture has to be changed instead. This work proposes a Human-Centered Deep Compositional model (HCDC) that combines low-level visual discrimination of a CNN and the high-level reasoning of a Hierarchical Compositional model (HCM). Defined as a transparent model, it can be optimized to real-world environments by adding compactly encoded domain knowledge from human studies and physical laws. The new FridgeNetv2 dataset and a mixture of publicly available datasets are used as a benchmark. The experimental results show the proposed model is explainable, has higher discriminative and generative power, and better handles the occlusion than the current state-of-the-art Mask-RCNN in instance segmentation tasks.

• Jezik: Angleški jezik
• Ključne besede: computer vision, deep learning, convolutional neural networks, hierarchical compositional model, occlusion, discriminability, generalizability, interpretability, domain knowledge, instance segmentation, occlusion handling
• Vrsta gradiva: Članek v reviji
• Tipologija: 1.01 - Izvirni znanstveni članek
• Organizacija: FE - Fakulteta za elektrotehniko
• Status publikacije: Objavljeno
• Različica publikacije: Objavljena publikacija
• Leto izida: 2023
• Št. strani: 14 str.
• Številčenje: Vol. 138, art. 109397
• PID: 20.500.12556/RUL-148675
• UDK: 004
• ISSN pri članku: 0031-3203
• DOI: 10.1016/j.patcog.2023.109397
• COBISS.SI-ID: 142438403
• Datum objave v RUL: 29.08.2023

Gradivo je del revije

• Naslov: Pattern recognition
• Skrajšan naslov: Pattern recogn.
• Založnik: Elsevier
• ISSN: 0031-3203
• COBISS.SI-ID: 26103040

Licence

• Licenca: CC BY-NC-ND 4.0, Creative Commons Priznanje avtorstva-Nekomercialno-Brez predelav 4.0 Mednarodna
• Povezava: http://creativecommons.org/licenses/by-nc-nd/4.0/deed.sl
• Opis: Najbolj omejujoča licenca Creative Commons. Uporabniki lahko prenesejo in delijo delo v nekomercialne namene in ga ne smejo uporabiti za nobene druge namene.

Sekundarni jezik

• Jezik: Slovenski jezik
• Ključne besede: računalniški vid, globoko učenje, konvolucijske nevronske mreže, hierarhični kompozicionalni model, zakrivanje, interpretabilnost, poznavanje področja