Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Next revision		Previous revision
highlights:2025:foramslice [2025/12/02 06:21] – created fits draft Didier Barradas Bautistahighlights:2025:foramslice [2025/12/02 06:38] (current) – Didier Barradas Bautista
Line 2: Line 2:
  
 <WRAP center> <WRAP center>
-<fs 28>Automated Fossil Species Identification from Micro-CT Scans Using Deep Learning</fs>+<fs 32>Automated Fossil Species Identification from Micro-CT Scans Using Deep Learning</fs>
 </WRAP> </WRAP>
  
 <WRAP group> <WRAP group>
 <WRAP third column> <WRAP third column>
 +
 <WRAP center> <WRAP center>
-{{:wiki:highlights:2025:foram_overview.png?400|}} +{{:wiki:highlights:2025:foram_overview.png?600|}}
-</WRAP>+
 </WRAP> </WRAP>
  
-<WRAP twothirds column>+<WRAP center round box 100% large> {{:favicon.ico?nolink&0x30}} <fs:26px>** KVL Staff on Project**</fs>
  
-<WRAP hi> +{{:icon:twbs:link:w-25:h-auto:person.svg?nolink&}} Abdelghafour Halimi \\ 
-<fs 20>Revolutionizing micropaleontology with AI!</fs> +{{:icon:twbs:link:w-25:h-auto:envelope-at.svg?nolink&}} abdelghafour.halimi@kaust.edu.sa \\
-</WRAP>+
  
-Foraminifera are microscopic organisms whose fossilized shells reveal Earth's climate history. **ForamDeepSlice** uses deep learning to automatically classify 12 foraminifera species from micro-CT scans with unprecedented accuracy.+{{:icon:twbs:link:w-25:h-auto:person.svg?nolink&}} Didier Barradas-Bautista \\ 
 +{{:icon:twbs:link:w-25:h-auto:envelope-at.svg?nolink&}} didier.barradas@kaust.edu.sa \\
  
-<WRAP important> +{{:icon:twbs:link:w-25:h-auto:person.svg?nolink&}} Ronell Sicat \\ 
-**🎯 Key Results:** +{{:icon:twbs:link:w-25:h-auto:envelope-at.svg?nolink&}} ronell.sicat@kaust.edu.sa \\
-  * **95.6%** accuracy across 12 species +
-  * **99.6%** top-3 accuracy  +
-  * **109,617** 2D slices from 97 specimens +
-  * Interactive dashboard for real-time classification +
-</WRAP>+
  
-</WRAP>+{{:icon:twbs:link:w-25:h-auto:building.svg?nolink&}} Building 1, Level 0, Office 0125 \\
 </WRAP> </WRAP>
  
-----+<WRAP center round box 100%> {{:wiki:kaust-seeds.png?nolink&0x30}} <fs:26px>** Collaborators**</fs>
  
-===== 📊 Research Overview ===== 
  
-<WRAP center> +{{:icon:twbs:link:w-25:h-auto:person.svg?nolink&}} Ali Alibrahim \\ 
-{{:wiki:highlights:2025:foram_pipeline.png?900|}} +{{:icon:twbs:link:w-25:h-auto:envelope-at.svg?nolink&}} ali.alibrahim@kaust.edu.sa \\
-</WRAP>+
  
-<WRAP group> +{{:icon:twbs:link:w-25:h-auto:person.svg?nolink&}} Abdulkader M. Afifi \\ 
-<WRAP half column> +{{:icon:twbs:link:w-25:h-auto:envelope-at.svg?nolink&}} abdulkader.alafifi@kaust.edu.sa \\
-<WRAP notice box 100%> +
-**Dataset** +
-  * 97 micro-CT scanned specimens +
-  * 27 species total, 12 selected for training +
-  * 109,617 high-quality 2D slices +
-  * Rigorous specimen-level data splitting +
-</WRAP> +
-</WRAP>+
  
-<WRAP half column> 
-<WRAP notice box 100%> 
-**Methods** 
-  * 7 state-of-the-art CNN architectures tested 
-  * Novel PatchEnsemble strategy 
-  * Transfer learning from ImageNet 
-  * Advanced data augmentation 
-</WRAP> 
-</WRAP> 
 </WRAP> </WRAP>
  
-==== Top Performing Models ====+</WRAP> /* end column */
  
-^ Model ^ Accuracy ^ F1-Score ^ +<WRAP twothirds column>
-| **ForamDeepSlice (Ensemble)** | **95.6%** | **95.0%** | +
-| ConvNeXt-Large | 95.1% | 94.1% | +
-| NASNet | 93.7% | 92.5% | +
-| ResNet101V2 (Baseline) | 84.3% | 80.8% |+
  
-Most species achieved F1-scores **exceeding 90%**, with the best reaching **99.7%**. Our PatchEnsemble approach significantly improved classification of challenging species.+===== Research Breakthrough ===== 
 +<WRAP> 
 +Foraminifera are microscopic organisms whose fossilized shells provide insight into the history of Earth's climate. Our **ForamDeepSlice** framework uses deep learning to automatically classify 12 foraminifera species from micro-CT scans with unprecedented accuracy. 
 +  * 📊 Tested on 97 specimens representing 27 species, ForamDeepSlice achieved **95.6%** accuracy and **99.6%** top-3 accuracy across 109,617 2D slices. 
 +  * 🎯 The method includes an interactive dashboard for real-time classification and 3D slice matching. 
 +  * 🔬 This work establishes new benchmarks for AI-assisted micropaleontological identification. 
 +</WRAP>  
 +  
  
----- 
- 
-===== 💻 Interactive Dashboard ===== 
  
 +<WRAP notice>
 +<fs 18>**🔬 Research Visualization**</fs>
 +The workflow diagram below illustrates our comprehensive deep learning pipeline for automated foraminifera classification. The paper entitled "ForamDeepSlice: A High-Accuracy Deep Learning Framework for Foraminifera Species Classification from 2D Micro-CT Slices" is available for download and review [[https://arxiv.org/abs/2512.00912|here]]. 
 +</WRAP>
 <WRAP center> <WRAP center>
-{{:wiki:highlights:2025:foram_dashboard.png?800|}}+{{:wiki:highlights:2025:foramdeepslice_workflow_v1.jpg?1000}}
 </WRAP> </WRAP>
  
-<WRAP group> +<fs 18>**🎯 KVL's Contribution**</fs> 
-<WRAP half column> +KVL's visualization scientists **Abdelghafour Halimi**, **Didier Barradas-Bautista**, and **Ronell Sicat** contributed significantly to the development of the ForamDeepSlice framework through:
-<color #000000> +
-**Classification Features:** +
-  * Drag-and-drop image upload +
-  * Automatic preprocessing & segmentation +
-  * Real-time species prediction +
-  * Confidence scores & rankings +
-  * Works with micro-CT and optical images +
-</color> +
-</WRAP>+
  
-<WRAP half column> +  * Design and implementation of PatchEnsemble strategy for improved classification accuracy 
-<color #000000> +  * Curation of a comprehensive micro-CT foram dataset with rigorous specimen-level splitting 
-**3D Slice Matching:** +  * Evaluation and benchmarking of 7 state-of-the-art CNN architectures 
-  * Find slice orientation in 3D models +  * Development of an interactive dashboard with real-time classification and 3D slice matching 
-  * Advanced similarity metrics (SSIM, NCC, Dice) +  * Optimization of preprocessing pipeline and data augmentation strategies 
-  * Interactive 3D visualization +  * Public release of the ForamDeepSlice framework for scientific community use
-  * No coding required +
-</color> +
-</WRAP> +
-</WRAP>+
  
-<WRAP center important> 
-[[https://youtu.be/betkJ3gsmNQ|📺 Watch Video Tutorial]] 
-</WRAP> 
  
----- +</WRAP> /* end column */ 
- +</WRAP> /* end group */
-===== 🌟 Impact & Applications ===== +
- +
-<color #000000> +
-**Scientific Contributions:** +
-  * First comprehensive micro-CT foram classification dataset +
-  * Accelerates fossil identification workflows (reduces expert time) +
-  * Supports biostratigraphy and climate reconstruction research +
-  * Fully reproducible Docker environment +
-  * Open-source framework for domain scientists +
- +
-**Future Applications:** +
-  * Expand to additional species +
-  * Mobile applications for field use +
-  * Integration with geological context +
-</color> +
- +
-<WRAP important> +
-**Note:** ForamDeepSlice is a decision-support tool for experts, complementing—not replacing—careful taxonomic analysis. +
-</WRAP> +
- +
----- +
- +
-===== 📚 Resources ===== +
- +
-<WRAP center tip box 80%> +
-**Project Team:** Abdelghafour Halimi, Didier Barradas-Bautista, Ronell Sicat (KVL) | Ali Alibrahim, Abdulkader M. Afifi (PSE Division) +
- +
-[[https://arxiv.org/|📖 Read Full Paper]] | [[https://github.com/|💻 GitHub Code]] | [[https://youtu.be/betkJ3gsmNQ|🎥 Video Demo]] +
-</WRAP>+
  
-**Acknowledgments:** KAUST Core Labs, Supercomputing Core Lab (Ibex), Domingo Lattanzi-Sanchez (Geo-Energy Platform) 
  
 ---- ----
-{{tag>highlight kvl fossils AI deep-learning paleontology}}+{{tag>highlight kvl}}