Trung-Nghia Le

A simple camera and a printed layout can replace traditional keyboards, making mobile computing more accessible and user-friendly.

Vision-language models can achieve over 96% accuracy in identifying dangerous actions by children, far surpassing traditional deep learning methods.

Achieving state-of-the-art image synthesis detection accuracy with a minimalist method that runs efficiently on low-end devices could revolutionize how we combat image-based deception.

MythraGen achieves a new benchmark in text-to-art generation by combining retrieval techniques with model fine-tuning, producing artworks that resonate deeply with user prompts.

Achieving substantial model compression with negligible accuracy loss could redefine deployment strategies for neural networks on edge devices.

SIFT-based image matching can accurately assess drawing skills, outperforming traditional evaluation methods and simplifying the assessment process.

Synthetic images generated by Stable Diffusion can elevate indoor scene recognition accuracy to 100%, even with lightweight models.

Current forensic methods fail dramatically against unseen synthetic disasters, revealing a critical vulnerability in our ability to discern reality from AI-generated fabrications.

Context-rich image captions are now possible by leveraging structured external knowledge, transforming how we understand image description generation.

CIAN boosts image captioning quality by integrating external narratives, achieving a CIDEr score increase from 0.030 to 0.094 on the OpenEvents-V1 benchmark.

Achieving artifact-free dynamic head editing, Edit3DGS enables seamless expression transformations and appearance refinements in real-time video.

Visual-RAG reveals that machines can mimic human-like creativity by transforming ambiguous silhouettes into compelling animal art, but still struggle with achieving high perceptual impact.