Jirince Biaba

Jirince Biaba

Auto-generated profile, based on publicly available information

Jirince Biaba

Hanoi University of Science and Technology

Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

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

Representative Publications

View All Publications
28 December 2025 in Analecta Technica Szegedinensia
Verified
Journal Article
Open access
FOFiston Oshasha
Fiston Oshasha
SDSaint Jean Djungu
Saint Jean Djungu
Alidor M. MbayandjambeAlidor M. Mbayandjambe
Alidor M. Mbayandjambe
FMFranklin Mwamba
Franklin Mwamba
JBJirince Biaba
Jirince Biaba
BMBlaise Muhala
Blaise Muhala
Atmospheric fog poses a critical challenge for computer vision systems in autonomous driving, surveillance, and robotics, where reliable object classification is essential. Under severe fog, classification accuracy can degrade by over 50%, and most existing approaches rely on separate defogging steps, which limit their applicability in real-time settings. This study introduces the Enhanced Density-Aware Cross-Scale Transformer (EDCST), a novel architecture designed for direct object classification under foggy conditions without requiring prior defogging. To support model training and evaluation, we developed a physics-based simulation framework generating four fog types (uniform, gradient, patchy, and adaptive) across nine intensity levels. EDCST leverages 384 dimensional embeddings, eight transformer layers, and twelve attention heads, trained using curriculum learning and OneCycleLR scheduling. On CODaN-Fog (15,500 images at 224×224 resolution), EDCST achieves 84.4% accuracy on clean images and retains 74.2% accuracy under severe fog (80% intensity), outperforming baseline transformers by 15.8%. Class-wise sensitivity analysis reveals that larger objects, such as vehicles and animals, maintain over 75% classification performance, while smaller objects are more affected. Patchy fog causes the greatest accuracy drop (19.1%), followed by adaptive (8.9%) and uniform fog (6.8%). The model converges in 100 epochs within 513 minutes. This work introduces a real-time-capable classification framework that eliminates defogging requirements and maintains strong performance under diverse fog conditions, making it highly suitable for safety-critical vision applications.

10 September 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
FOFiston Oshasha
Fiston Oshasha
FMFranklin Mwamba
Franklin Mwamba
JBJirince Biaba
Jirince Biaba
SDSaint Jean Djungu
Saint Jean Djungu
Epidemic situation reports (SitRep) frequently publish a national total alongside tables disaggregated by health zone. These two administrative layersdo not always coincide, which complicates operational interpretation andmay conceal local incompleteness in reporting. Drawing on publicly releasedSitReps from the 2026 Bundibugyo virus disease (BDBV) outbreak in theDemocratic Republic of the Congo, this study quantifies national-versuszonal discrepancies, estimates a reporting-completeness proxy π and the complementary under-reporting measure 1-π, and compares a hierarchical deepmodel with classical baselines without relying on a nominative case linelist.A multi-metric reconciliation panel is constructed from open INSP SitReps; ahierarchical network yields latent allocation shares λ and zone-level completeness π; residual allocation is assessed through leave-zone-out validation andis contrasted with population-proportional allocation, iterative proportionalfitting (IPF), and empirical-Bayes shrinkage; bootstrap confidence intervalsand a controlled simulation complete the evaluation. For confirmed cases, the mean overlap gap remains moderate (+4.2), while the mean under-reportingproxy is low (0.0106; 95% CI [0.0099, 0.0112]). Under leave-zone-out evaluation, IPF improves residual allocation by approximately 85% relative to thepopulation baseline, and the deep model by approximately 73%. Suspectedcase series prove fragile owing to sparse zonal coverage. Overall, national–zonal SitRep reconciliation provides a practical diagnostic of data qualityand a reporting proxy even in the absence of a linelist. Classical methodsremain competitive for residual allocation, whereas the deep approach principally contributes covariate-informed allocation and a prioritization map foroperational review.