GitHub Projects

 

I maintain a range of open-source repositories that reflect my research in AI, robotics, computer vision, and ecological monitoring. These projects span from ecological datasets and conservation tools to advanced deep learning models and bio-inspired robotics systems.

🟢 Ecological AI & Conservation

• EcoDetectX – A generalized detection and evaluation platform for ecological research. Supports YOLO and other deep learning models, enabling single-image and batch folder detection, annotation export, and automated CSV reports for large-scale conservation studies.

• AutoCoralMatch – A full pipeline for automated coral bleaching assessment. Integrates underwater dehazing, object detection (YOLOv12), segmentation (SAM2), CoralWatch card rectification, and perceptual ΔE scoring for standardized bleaching severity analysis.

• AI-CoralWatch – Dedicated AI pipeline for detecting CoralWatch reference cards in field imagery, enabling accurate patch-level color analysis.

• AI-CoralHealth – A framework for automated coral health classification, leveraging color metrics, patch segmentation, and expert-aligned scoring methods.

🟢 Bio-Inspired Robotics

• HuBot – A biomimetic robotic platform replicating the morphology and behavior of the Houbara bustard. Designed for ecological studies, conservation trials, and kinetic art installations.

• HuBotDetSeg – Specialized detection and segmentation models for monitoring HuBot in experimental field deployments, supporting synchronized multi-camera analysis.

• Kinebot – Bio-Inspired Robotic Houbara: From Development to Field Deployment for Behavioral Studies.

• mobile-houbara-detseg – Lightweight, mobile-optimized models for real-time detection and segmentation of Houbara in field environments with limited computational resources.

🟢 Benchmark Datasets & Evaluation

• FishDet-M – A merged benchmark dataset containing over 100k annotated fish images from multiple sources, supporting robust evaluation of underwater detection models.

• EvalDet – A flexible framework for evaluating detection models across multiple datasets, with metrics such as mAP, IoU, and per-class breakdowns.

• FishDet-UI – A web-based visualization and annotation interface for exploring and refining large underwater detection datasets.

🟢 Core AI & Vision Models

• KINETIQA – Transformer-based architecture for real-time knee joint angle prediction in exoskeleton-assisted gait rehabilitation. Features temporal convolution blocks, recursive feature elimination, and quality-driven attention mechanisms.

• GFN – Gated Fusion Network for underwater image dehazing, combining multi-scale attention and feature fusion to restore clarity and color fidelity in turbid imagery.

• GFN_dehazing – An optimized and extended version of GFN with modular components for benchmarking underwater image enhancement tasks.

• VLAs – Vision-Language-Action models adapted for robotic manipulation, enabling multimodal reasoning across visual cues, language commands, and action planning.

• VisualSLAM – Experiments integrating semantic perception into SLAM for robust vision-based navigation and mapping.

• SAFARI – An autonomous robotics framework for field deployment in ecological and conservation monitoring, with support for multi-sensor integration and adaptive behaviors.

👉 Explore all repositories here: https://github.com/LyesSaadSaoud