Mohamed Jalal BAIM

A novel editing method built on Stable Diffusion 3 that enables precise, localized edits using only text prompts. It automatically identifies which regions to modify via velocity field analysis, then applies masked velocity blending to confine edits to the relevant area.

An image editing pipeline that understands complex natural language instructions and applies precise, context-aware edits. Using a multi-stage pipeline: text → plan generation → object grounding → edit → validation.

A novel method to generate interpretable counterfactual explanations for image classification models using prototype-based guidance. The method combines autoencoders with prototype learning to produce plausible and class-representative counterfactual examples.

Trustworthy Retrieval-Augmented Generation for the medical domain

A medical-domain QA system built on hybrid dense + sparse retrieval with hallucination prevention. Evaluated with RAGAS: Faithfulness, Answer Relevancy, Context Precision.