By the research group of Associate Prof. Dr. Sukalpa Chanda, Østfold University College (HiØ)

Research activities at Østfold University College (HiØ), under the supervision of Associate Prof. Dr. Sukalpa Chanda, continue to advance the frontiers of computer vision, multimodal artificial intelligence, generative modeling, and intelligent document analysis. Recent research outcomes include two papers accepted at ECCV 2026, a publication in Transactions on Machine Learning Research (TMLR), a journal article in IJDAR, and multiple contributions to ICDAR 2025.

The two ECCV 2026 papers highlight emerging research directions in AI-driven visual understanding. FUSE introduces a physics-guided framework for non-contact blood oxygen (SpO₂) estimation from facial videos [1], while SpEmoC addresses multimodal emotion recognition through a balanced benchmark integrating visual, acoustic, and textual information [2]. These contributions demonstrate the group's growing interest in physics-informed learning and human-centered multimodal AI.

In video understanding, the TMLR paper, T2L, explores efficient zero-shot action recognition through temporal token learning, advancing the adaptability of vision-language models [3]. Complementing this work, CA-DIFF, published in IJDAR, investigates diffusion-based handwriting generation with improved stylistic consistency [4]. The group's ICDAR 2025 contributions further explore writer analysis, handwriting synthesis, and scene text generation using graph-based learning and generative approaches [5, 6].

Together, these achievements reflect the ongoing research and doctoral supervision activities at HiØ, fostering international collaborations and advancing innovative AI methodologies. The research group under the supervision of Dr. Sukalpa Chanda continues to investigate robust, efficient, and generalizable learning approaches, connecting fundamental advances in artificial intelligence with real-world applications in healthcare, human-centered computing, and document intelligence.

[1] S. Ahmad et al., FUSE: Filter-Free Unified Spatiotemporal Estimation of SpO₂ via Wave-Transport Modeling, ECCV 2026.

[2] S. Bano et al., SpEmoC: A Balanced Speaker-Segment Multimodal Emotion Benchmark, ECCV 2026.

[3] S. Ahmad et al., T2L: Efficient Zero-Shot Action Recognition with Temporal Token Learning, Transactions on Machine Learning Research 2025.

[4] A. Gurav et al., CA-DIFF: Enhancing Stylistic Consistency in Handwriting Generation—A Diffusion Model Perspective, International Journal on Document Analysis and Recognition 2026.

[5] A. Gurav et al., Beyond Memorization: Training-Free Style Mixing for Variability in Handwritten Text Generation Using Writer Embedding Injection in Pretrained Diffusion Models, ICDAR 2025.

[6] K. Priya et al., Graph Convolutional Teacher-Student Framework for Writer Inspection from Intra-variable Handwritten Words, ICDAR 2025.