Enabling Perspective-Aware Ai with Contextual Scene Graph Generation by Platnick, Daniel, Alirezaie, Marjan, Rahnama, Hossein

“…We evaluated PASGG-LM pipelines using state-of-the-art SGG models, including Motifs, Motifs-TDE, and RelTR, and showed that fine-tuning LLMs, particularly GPT-4o-mini and Llama-3.1-8B, improves performance in terms of R@K, mR@K, and mAP. Our method is capable of generating scene graphs that capture complex contextual aspects, advancing human–machine interaction by enhancing the representation of diverse perspectives. …”
Get full text

Decoding Surface Touch Typing from Hand-Tracking by Richardson, Mark, Durasoff, Matt, Wang, Robert

Get full text

Semantic sensitive satellite image retrieval by Li, Yikun

“…The main difficulty is that users think in terms of high-level semantics, which cannot be directly mapped onto extracted low-level features / signals. …”
Get full text

Tripartite partnership for the transformation of the Singapore maritime industry - Part B : Industry's perspective by Tan, Xian Jun

“…In responses to the threat of the changing economy, the Singapore government have launched several initiatives including the Industry Transport Maps (ITMs) to facilitate an effective tripartite partnership towards the transformation of Singapore’s Maritime Industry. …”
Get full text

Imaging and processing of tree trunk by Hong, Yuan Ming

“…Therefore, further investigations were conducted to better understand the peak time differences between 2D simulation and 3D simulation where the results can help to map the 2D simulations into 3D simulation using machine learning.…”
Get full text

Deep visual saliency on stereoscopic images by Nguyen, Anh-Duc, Kim, Jongyoo, Oh, Heeseok, Kim, Haksub, Lin, Weisi, Lee, Sanghoon

“…Our results from thorough experiments confirm that the predicted saliency maps are up to 70% correlated with human gaze patterns, which emphasize the need for the hand-crafted features as input to deep neural networks in S3D saliency detection.…”
Get full text

Ultra-low power real-ime object detection based on quantized CNNs by Chew, Jing Wei

“…Using this approach, the proposed model can run at 37.7 FPS on an NVIDIA Jetson Nano with a peak memory footprint of 17.1 MB, while attaining a reasonable mAP@0.50 Intersection over Union (IoU) of 0.37 on the Pascal Visual Object Classes (VOC) dataset. …”
Get full text

Multi-view human action recognition in meeting scenarios by Yin, Haixiang

“…The method proposed in this project can improve by up to nearly 10 percent of the mAP (Mean Average Precision) on AMI meeting corpus compared to single-view recognition approaches.…”
Get full text

Visual place recognition for autonomous robots using deep learning by Huang, Yifeng

“…In autonomous and unmanned aerial vehicles, visual location recognition helps to detect actual destinations and locations In order to match standard features from different angles of a scene, we propose a global feature matching method, which is separate from the current popular local feature matching method. We build attention maps from two dimensions to refine the calculation methods of residuals in VLAD, channel and spatial, and get better results than the original NetVLAD model. …”
Get full text

Mobile robot localization in geometrically symmetric environments by Lyu, Yaozeng

“…In addition, the similar magnetic field single fingerprint information can easily cause mismatches in the magnetic field map, then lead to localization failures. This project will explore and propose novel magnetic field-based localization algorithms for mobile robots to achieve highly robust and precise localization in geometrically symmetric environment.…”
Get full text

Thermodynamic properties of Xenon storage system under adsorption condition by Koh, Darren Shi Hao

“…Furthermore, through the entropy and temperature maps analysis, a typical heat storage system can be represented where the charging and discharging of Xenon (Xe) gas is represented by the change of entropy in a closed loop.…”
Get full text

Hydrological & hydrodynamic modelling of the Baliem River in Highland Papua, Indonesia by Wee, Khai Zher

“…We first delineated the watersheds for these pour points in ArcMap to characterise their attributes. Thereafter, we generated simulated river discharges at these pour points with meteorological data in the Génie Rural à 4 paramètres Journalier (GR4J) rainfall-runoff hydrological model. …”
Get full text

Deep neural network model for hypertension risk level detection using ECG signal by Lau, Kai Yun

“…The deep convolutional network consists of 1D convolution and max pooling layers, which extract feature mappings and perform risk level classifications. It classifies the ECGs into ‘normotensive’ (no risk), ‘hypertension’ (low risk), and ‘hypertension + any recorded event within 12 months of recording’ (high risk). …”
Get full text

Autonomy mobile or progressive web app for elderly by Liu, Fenghao

“…These reminders can include important tasks such as scheduling physical examinations, remembering to take medications on time, and even helping with navigating maps. By incorporating these functionalities, the software aims to simplify daily routines for seniors and promote their overall well-being and independence. …”
Get full text

Deposition characterisation and process window development for plasma sprayed thermal barrier coatings by Ng, Heong Wah.

“…Generally, process control is to ultimately control the temperature and velocity of particles which impact on the substrate. A process map which indicate the best control location for optimum operation is derived by simulation and experimentation in the paper.…”
Get full text

Estimation of traffic flow using sparse camera images from LTA by Ong, Justin Yang Chen.

“…The processed images along with its assigned road condition types are then displayed on the web browser with Google Maps. Three algorithms namely, (1) Manual, (2) Percentile-based and (3) K-nearest neighbours, were used and experimented to find out which algorithm provides the highest accuracy in estimating and assigning the correct road condition type to a test image. …”
Get full text

Development of a repair service locator Android application by Jiang, Jessie Jielin

“…Moreover, the application is able to link to another travel application, Google Maps, to get directions. Hence, users can easily get directions to travel to the service center. …”
Get full text

Characterizing the spectrum of slip behavior at the Sumatran subduction zone with long-term geodetic records by Tsang, Louisa Lok Hang

“…Our results highlight the need for researchers to treat interseismic fault coupling maps as a snapshot of conditions on the fault that can change with time. …”
Get full text

Big data analytics for smart transportation (2) by Chung, Maybelle Huishi

“…The results from the clustering was visualized on a map of Singapore and the travel patterns were identified. …”
Get full text

‘You could start a new country, but you would need to discover new land’: exploring understandings of statehood with primary school pupils in England by Saddington, L, McConnell, FR

“…Geographical education in schools draws upon a range of representations from atlases to wall maps, which portray a highly ordered image of a world that is neatly divided up into bounded nation-states. …”