Open to opportunities

Roua Kammoun

@rouakammoun

Message

I build computer vision and multimodal AI systems for sign language and medical imaging.

France

Message

What I'm looking for

I'm looking to contribute to computer vision and multimodal AI work, especially spatiotemporal learning, vision-language models, and real-time AI systems.

I'm building computer vision and multimodal AI systems for gesture recognition, sign language, and medical imaging. At XLIM Laboratory, University of Limoges, I designed a medical gesture recognition system using RGB and skeleton video data, ST-GCNs, Transformers, and vision-language models.

I applied few-shot learning for cross-signer generalization and achieved 90.3% signer-independent accuracy on the KaRSL dataset, surpassing prior baselines. I also evaluated performance under occlusion, lighting variation, and inter-signer variability through detailed failure-mode analysis.

Previously at the Digital Research Center of Sfax, I built a multimodal Arabic Sign Language recognition system that reached 99% signer-dependent and 83% signer-independent accuracy. At ZNet-IT, I developed an Android-deployed skin lesion classifier covering 34 dermatological conditions with 98% accuracy.

I'm completing a Computer Engineering degree focused on AI, and I enjoy turning research into real-time products such as a sign-language translator, an interview RAG assistant, and a TikTok trends dashboard.

Experience

Work history, roles, and key accomplishments

Computer Vision Research Intern

XLIM Laboratory, University of Limoges

Feb 2026 - Jun 2026 (4 months)

Designed a medical gesture recognition system from multimodal video data, combining sequential modeling with Vision-Language Models. Achieved 90.3% signer-independent accuracy on the KaRSL dataset and evaluated robustness under various conditions.

Computer Vision Deep Learning ST GCN Transformers Vision Language Models Few Shot Learning Multimodal Fusion PyTorch OpenCV Mediapipe

Computer Vision Research Intern

Digital Research Center of Sfax (CRNS)

Jun 2025 - Aug 2025 (2 months)

Built a multimodal Arabic Sign Language recognition system using ST-GCNs, Transformers, and CNNs. Achieved 99% signer-dependent and 83% signer-independent accuracy, outperforming prior state-of-the-art models.

Computer Vision Deep Learning ST GCN Transformers CNN Multimodal Fusion PyTorch TensorFlow OpenCV

Mobile & AI Development Intern

ZNet-IT

Jun 2024 - Aug 2024 (2 months)

Developed a skin lesion classification system detecting 34 conditions with 98% accuracy via a fine-tuned CNN deployed on Android. Optimized the inference pipeline for mobile deployment, reducing latency while preserving accuracy.

Deep Learning CNN Flask Android Python OpenCV TensorFlow Keras