Visual Relationship Detection with AI - Versionsgeschichte

Moye-nyuysoni.glein-perry@stud.hshl.de am 14. April 2025 um 15:17 Uhr

2025-04-14T15:17:26Z

← Nächstältere Version		Version vom 14. April 2025, 15:17 Uhr
Zeile 1:		Zeile 1:
	~~<!--~~
	{\| class="wikitable"		{\| class="wikitable"
	\|-		\|-
Zeile 12:		Zeile 11:
	\| '''Betreuer''': \|\| [[Benutzer:Ulrich_Schneider\| Prof. Dr.-Ing. Schneider]]		\| '''Betreuer''': \|\| [[Benutzer:Ulrich_Schneider\| Prof. Dr.-Ing. Schneider]]
	\|}		\|}
	~~-->~~
	= Abstract =		= Abstract =
	Visual Relationship Detection (VRD) enables machines to go beyond object detection and recognize how objects interact within an image. This project combines a pre-trained YOLOv2 deep learning model in MATLAB with heuristic rules to detect object-object relationships, such as “person riding a bicycle” or “dog under table.” Additionally, we evaluate multiple VRD 2 approaches using a Zwicky Box based on technical criteria like accuracy, interpretability, and computational cost. The project is fully documented in a wiki, supported by visualizations.		Visual Relationship Detection (VRD) enables machines to go beyond object detection and recognize how objects interact within an image. This project combines a pre-trained YOLOv2 deep learning model in MATLAB with heuristic rules to detect object-object relationships, such as “person riding a bicycle” or “dog under table.” Additionally, we evaluate multiple VRD 2 approaches using a Zwicky Box based on technical criteria like accuracy, interpretability, and computational cost. The project is fully documented in a wiki, supported by visualizations.

2025-04-11T23:35:29Z

Research on Solutions: Morphological Box

← Nächstältere Version		Version vom 11. April 2025, 23:35 Uhr
Zeile 45:		Zeile 45:

	=Research on Solutions: Morphological Box=		=Research on Solutions: Morphological Box=

			{\| class="wikitable"
			\|+
			\|-
			! Model Type !! Training Required!! RequiresData !! Development Effort !! Accuracy
			\|-
			\| Pretrained \|\| No \|\| No\|\| Low \|\| Meduim
			\|-
			\| TrainfromScratch\|\| Yes\|\| Yes \|\| Meduim to High \|\| Often Lower
			\|}

			'''Model Type'''

			Pretrained Model: A model like YOLO or ViLD that is already trained on large datasets (e.g., COCO).

			'''Train from Scratch:'''

			Build your own model and train it from zero using labeled VRD data.

			'''Training Required'''

			Whether the model needs you to train it before using.

			'''Requires Data'''

			Indicates if a large labeled dataset is needed to make the system work.


			'''Development Effort'''

			Time and complexity involved in building and testing the system.

			Pretrained = plug-and-play; from scratch = architecture + dataset + training + validation.

			'''Accuracy'''

			Pretrained models are reliable for common objects/relationships but may miss fine details.

			From-scratch models can start weak and only improve with massive data and tuning

	=System Overview and Code Explanation=		=System Overview and Code Explanation=

2025-04-11T23:27:27Z

Introduction

← Nächstältere Version		Version vom 11. April 2025, 23:27 Uhr
Zeile 43:		Zeile 43:

	Documents all results in a structured scientific format.		Documents all results in a structured scientific format.

			=Research on Solutions: Morphological Box=

	=System Overview and Code Explanation=		=System Overview and Code Explanation=

2025-04-11T23:22:47Z

Required Toolboxes / Add-Ons

← Nächstältere Version		Version vom 11. April 2025, 23:22 Uhr
Zeile 19:		Zeile 19:
	You’ll need the following MATLAB add-ons installed:		You’ll need the following MATLAB add-ons installed:

	Computer Vision Toolbox		'''Computer Vision Toolbox'''

	Needed for object detection functions, bounding box operations, and detect().		Needed for object detection functions, bounding box operations, and detect().

	Deep Learning Toolbox		'''Deep Learning Toolbox'''

	Required for working with neural networks, including YOLO.		Required for working with neural networks, including YOLO.

	Deep Learning Toolbox Model for YOLO v2 Object Detection		'''Deep Learning Toolbox Model for YOLO v2 Object Detection'''

	This is the pretrained YOLO v2 model you are using (tiny-yolov2-coco).		This is the pretrained YOLO v2 model you are using (tiny-yolov2-coco).

2025-04-11T23:20:47Z

Abstract

@@ Zeile 15: / Zeile 15: @@
 = Abstract =
 Visual Relationship Detection (VRD) enables machines to go beyond object detection and recognize how objects interact within an image. This project combines a pre-trained YOLOv2 deep learning model in MATLAB with heuristic rules to detect  object-object relationships, such as “person riding a bicycle” or “dog under table.” Additionally, we evaluate multiple VRD  2 approaches using a Zwicky Box based on technical criteria like accuracy, interpretability, and computational cost. The project is fully documented in a wiki, supported by visualizations.
 = Introduction =

2025-04-11T23:14:39Z

Testing & Proof of Functionality

← Nächstältere Version		Version vom 11. April 2025, 23:14 Uhr
Zeile 190:		Zeile 190:
	Explanation:		Explanation:
	We save the output showing the image and its relations		We save the output showing the image and its relations


	=Testing & Proof of Functionality=		=Testing & Proof of Functionality=
Zeile 197:		Zeile 198:


	[[Datei:PersonridingHorse.png\|~~1000px~~\|frameless\|centered]]		[[Datei:PersonridingHorse.png\|700px\|frameless\|centered]]
			[[Datei:PersonrdingBicycles.png\|700px\|frameless\|centered]]

	=Scientific Documentation=		=Scientific Documentation=

2025-04-11T23:11:58Z

Testing & Proof of Functionality

← Nächstältere Version		Version vom 11. April 2025, 23:11 Uhr
Zeile 197:		Zeile 197:


	~~Visual proofs (annotated outputs) were generated and included in the wiki~~.		[[Datei:PersonridingHorse.png\|1000px\|frameless\|centered]]

	=Scientific Documentation=		=Scientific Documentation=

2025-04-11T23:09:08Z

System Overview and Code Explanation

2025-04-11T22:57:27Z

Testing & Proof of Functionality

← Nächstältere Version		Version vom 11. April 2025, 22:57 Uhr
Zeile 114:		Zeile 114:

	=Testing & Proof of Functionality=		=Testing & Proof of Functionality=
	The system was tested with multiple static images (people, objects~~, animals~~).		The system was tested with multiple static images (people, objects).

	Correct object detection and relationship inference were observed.		Correct object detection and relationship inference were observed, but not every time.

	~~Edge cases (e.g., overlapping objects or occlusions) revealed the limitations of rule-based logic.~~

	Visual proofs (annotated outputs ~~and animated GIFs~~) were generated and included in the wiki.		Visual proofs (annotated outputs) were generated and included in the wiki.

	=Scientific Documentation=		=Scientific Documentation=

2025-04-11T22:27:16Z

System Overview and Code Explanation

@@ Zeile 111: / Zeile 111: @@
 Explanation:
 The unique relationships are collected and displayed in a GUI table for quick interpretation and documentation. This helps in validating system output.
 =Task=