Daniel Adu-Gyamfi

Daniel Adu-Gyamfi


Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
30 September 2024 in Information
Verified
Journal Article
Open access
PDF file
EGEnoch Opanin Gyamfi
Enoch Opanin Gyamfi
ZQZhiguang Qin
Zhiguang Qin
JDJuliana Mantebea Danso
Juliana Mantebea Danso
DADaniel Adu-Gyamfi
Daniel Adu-Gyamfi
Graph Neural Networks (GNNs) have gained popularity in image matching methods, proving useful for various computer vision tasks like Structure from Motion (SfM) and 3D reconstruction. A well-known example is SuperGlue. Lightweight variants, such as LightGlue, have been developed with a focus on stacking fewer GNN layers compared to SuperGlue. This paper proposes the h-GNN, a lightweight image matching model, with improvements in the two processing modules, the GNN and matching modules. After image features are detected and described as keypoint nodes of a base graph, the GNN module, which primarily aims at increasing the h-GNN’s depth, creates successive hierarchies of compressed-size graphs from the base graph through a clustering technique termed SC+PCA. SC+PCA combines Principal Component Analysis (PCA) with Spectral Clustering (SC) to enrich nodes with local and global information during graph clustering. A dual non-contrastive clustering loss is used to optimize graph clustering. Additionally, four message-passing mechanisms have been proposed to only update node representations within a graph cluster at the same hierarchical level or to update node representations across graph clusters at different hierarchical levels. The matching module performs iterative pairwise matching on the enriched node representations to obtain a scoring matrix. This matrix comprises scores indicating potential correct matches between the image keypoint nodes. The score matrix is refined with a ‘dustbin’ to further suppress unmatched features. There is a reprojection loss used to optimize keypoint match positions. The Sinkhorn algorithm generates a final partial assignment from the refined score matrix. Experimental results demonstrate the performance of the proposed h-GNN against competing state-of-the-art (SOTA) GNN-based methods on several image matching tasks under homography, estimation, indoor and outdoor camera pose estimation, and 3D reconstruction on multiple datasets. Experiments also demonstrate improved computational memory and runtime, approximately 38.1% and 26.14% lower than SuperGlue, and an average of about 6.8% and 7.1% lower than LightGlue. Future research will explore the effects of integrating more recent simplicial message-passing mechanisms, which concurrently update both node and edge representations, into our proposed model.

21 January 2022 in International Journal of System Assurance Engineering and Management
Verified
Journal Article
Closed access
PDF file
DADaniel Adu-Gyamfi
Daniel Adu-Gyamfi
AAAlbert Kofi Kwansah Ansah
Albert Kofi Kwansah Ansah
GAGabriel Kofi Armah
Gabriel Kofi Armah
SASeth Alornyo
Seth Alornyo
DADominic Kwasi Adom
Dominic Kwasi Adom
FZFengli Zhang
Fengli Zhang

5 January 2021 in Wireless Networks
Verified
Journal Article
Closed access
PDF file
DADaniel Adu-Gyamfi
Daniel Adu-Gyamfi
FZFengli Zhang
Fengli Zhang
ATAugustine Takyi
Augustine Takyi

29 September 2020 in BMC Medical Informatics and Decision Making
Verified
Journal Article
Open access
PDF file
DADaniel Adu-Gyamfi
Daniel Adu-Gyamfi
FZFengli Zhang
Fengli Zhang
AAAlbert Kofi Kwansah Ansah
Albert Kofi Kwansah Ansah
Background: A pandemic affects healthcare delivery and consequently leads to socioeconomic complications. During a pandemic, a community where there lives an asymptomatic patient (AP) becomes a potential endemic zone. Assuming we want to monitor the travel and/or activity of an AP in a community where there is a pandemic. Presently, most monitoring algorithms are relatively less efficient to find a suitable solution as they overlook the continuous mobility instances and activities of the AP over time. Conversely, this paper proposes an EDDAMAP as a compelling data-dependent technique and/or algorithm towards efficient continuous monitoring of the travel and/or activity of an AP. Methods: In this paper, it is assumed that an AP is infected with a contagious disease in which the EDDAMAP technique exploits a GPS-enabled mobile device by tagging it to the AP along with its travel within a community. The technique further examines the Spatio-temporal trajectory of the AP to infer its spatial time-bounded activity. The technique aims to learn the travels of the AP and correlates them to its activities to derive some classes of point of interests (POIs) in a location. Further, the technique explores the natural occurring POIs via modelling to identify some regular stay places (SP) and present them as endemic zones. The technique adopts concurrent object feature localization and recognition, branch and bound formalism and graph theory to cater for the worst error-guaranteed approximation to obtain a valid and efficient query solution and also experiments with a real-world GeoLife dataset to confirm its performance. Results: The EDDAMAP technique proofs a compelling technique towards efficient monitoring of an AP in case of a pandemic. Conclusions: The EDDAMAP technique will promote the discovery of endemic zones and hence some public healthcare facilities can rely on it to facilitate the design of patient monitoring system applications to curtail a global pandemic.

3 November 2021 in Journal of Information Technology Research
Verified
Journal Article
Open access
DADaniel Adu-Gyamfi
Daniel Adu-Gyamfi
FZFengli Zhang
Fengli Zhang

8 September 2020
Verified
Book Chapter
Closed access
PDF file
DADaniel Adu-Gyamfi
Daniel Adu-Gyamfi
FZFengli Zhang
Fengli Zhang
FZFan Zhou
Fan Zhou