Hi there 👋

Hi, I am Ekin from Istanbul. I am a Machine Learning Engineer (actually I have a bachelor's degree in Electronics and Communication Engineering). In general, my research topics are MLOps/AIOps and Evolutionary Algorithms (genetic algorithm, artificial bee colony algorithm etc.).

I have been interested in image processing since 2020, and machine learning since 2021. It is my code portfolio.

Table of contents

• Machine Learning (projects used scikit-learn, XGBoost, Catboost etc.)
• Deep Learning Classification (projects used Tensorflow with CNNs and ViTs models)
• Deep Learning Segmentation (UNet, DeepLabv3+, detectron2 etc. models)
• Deep Learning Object Detection (detectron2, YOLO etc. models)
• Natural Language Processing (text classification, topic modelling, prompting etc.)
• Others DL project (multiple instance learning, image captioning etc.)
• Hybrid Models (projects used a deep feature extractor and an ML classifier)
• Online/Incremental Learning (projects used online learning methods and libraries like River)
• Theoretical Machine Learning (some proofs of machine learning theory with python scripts)
• Custom Tensorflow Callbacks (for example, a callback creating a test set classification report during training)
• React Web App Projects (my React, Vite and Tailwind CSS projects)

Some of my projects

1. Machine Learning Projects

Gamma/Hadron Separation w/XGBoost, LightGBM, SVM (ROC AUC Score: 0.89)

Bears vs Pandas SVM, XGB, LGBM, Ensemble Method w/Noised-Dataset (ACC: 96.75 %)

Groundwater Quality Classification w/CatBoost

(QSAR) Prediction Biological Activity w/CatBoost Implementation (MCC: 0.825)

(QSAR) Classification Activity of Inhibitors of EphA4 Receptor Antagonists (AID 689) w/CatBoost (MCC: 0.782)

Alive/Dead Diabetic Outcome Prediction with CatBoostClassifier w/CatBoost (ROC AUC Score: 0.77)

Pistachio Classification w/CatBoost (ROC AUC Score: 0.9)

Rain Prediction w/CatBoost (F1-Score:0.84)

---

2. Deep Learning Projects

2.1 Classification

Endoscopy Image Classification w/Swin Transformer (F1 Score: 0.87)

Crop Disease Classification w/Feature Fusion (EfficientNet, MobileNet) (F1 Score: 0.8)

30 Plants Detection w/Custom ConvMixer based CNN (F1 Score: 0.77)

Bladder Tissue Classification w/ViT (F1 Score: 0.82)

Blood Cells Classification w/Custom ConvMixer based CNN (F1 Score: 0.98)

Brain Tumor Classification (Normal, Glioma, Meningioma, Pituitary) (Test ACC: 86.27 %)

Car Model and Color Multiclass Classification (F1 Score: 0.74)

Dental X-Rays Classification w/TPU (F1 Score: 0.72)

The Fashion Mnist Distributed DL Example (Mirrored Strategy) (Test ACC: 88.35 %)

Down Syndrome Detection w/CNN (Test AUC Score: 0.87)

Earthquake Events Classification (Mojor Event/Non-Major Event) (Test ACC: 64 %)

Fungus Detection w/10 Kfold CV Custom ConvMixer (F1 Score: 0.85)

Hieroglyph Multiclass Classification DenseNet (F1 Score: 0.85)

Higgs/Background Process Classification w/CNN using TPU (AUC Score: 0.83)

Jellyfish Classification (10KFold CV w/Custom ConvMixer) (F1 Score: 0.87)

MRI Sequence (T1, T2, T1 C+) Classification w/Custom CNN

Normal heartbeat/Myocardial Infarction Classification (ROC AUC Score: 0.842)

Pneumonia Detection w/Ensemble DL (Test AUC Score: 0.91)

Zipper Defect Classification (AUC Score: 0.98)

Glaucoma Classification w/ViT (F1 Score: 0.91)

Chest X-Ray Classification w/ViT (F1 Score: 0.9)

Document Classification w/ViT

(76 GB) 160 Polish Bird Sounds Classification

Rice Classification w/Custom ResNet50 (ACC 85%)

Maize Classification w/Custom ResNet18 (AUC Score: 0.98)

Sport Scene Classification w/ViT (3 KFold CV)

2.2 Segmentation

Brain tissue segmentation with U-net using TPU (Val Dice Coef: 0.88)

Brain tumor/anomaly segmentation with U-net using TPU

Asphalt Pavement Crack Segmentation U-Net

Eye Conjunctiva Segmentation with U-Net using TPU (Dice : 0.91, Jaccard : 0.82)

Iris Segmentation U-net w/TPU (Dice : 0.94, Jaccard : 0.88)

Particle Segmentation Custom DeepLabv3+ (Dice : 0.93, Jaccard : 0.88)

Retina Vessel Segmentation U-net w/TPU (Dice : 0.75)

Lung Segmentation UNet w/SeparableConv (Dice : 0.93)

Segmentation Medical Instrument w/Custom DeepLabv3+ (Dice : 0.86)

Tomato Segmentation w/detectron2 (mask AP: 61.94)

Brain Tumor Segmentation w/detectron2 (mAP@50:76.2)

Road Segmentation w/DeepLabv3+ from Scratch

Segmenting HuTu Cells DeepLabv3+ (Test Dice: 0.93)

2.3 Object Detection

Damaged Lamp Detection w/detectron2 (Faster R-CNN)

Plate Detection w/detectron2 (mAP@75: 89.19)

Tomato Detection w/detectron2 (mAP@50: 82.02)

Tiny Vehicle Detection w/detectron2 (mAP@50: 32.08)

Traffic Signs Detect w/detectron2 (mAP@50: 71.62)

Sign Detection w/Keras YOLO V8

Road Mark Detection (ResNet-50, ResNeXt 101 FPNs)

Bone Fracture Detection (ResNet-50, ResNeXt 101 FPNs)

Acne Detection w/Keras YOLO V8

Brain Tumor Detection w/Keras YOLO V8

2.4 Natural Language Processing

PaliGemma 3B for License Plate OCR

arXiv Articles Multi Label Classification w/FNet

IEEE Research Papers Topic Modelling w/LDA

Depressive vs Non-depressive Tweet w/Custom FNet (F1 Score: 0.88)

Yelp Review Stars Prediction (Classification) w/Gemma 7B (LoRA)

BBC News Topic Modeling w/LDA

Manufacturing Question-Answer w/Gemma 7B (Fine-Tuning LoRA) (Cosine Sim: 0.83)

Graph to Table w/Google's DePlot Model

Disease Article Topic Modelling w/BERTopic

Spam vs Ham Message w/ Gemma 7B Fine-Tuning (LoRA)

News Zero-Shot Topic Modelling w/BERTopic

Social Media Post Multiclass Classification w/DistilBERT

Gemma 2B Text Summarization w/Zero-Shot Prompting

Rating Prediction w/SentenceTransformer, CatBoost (MAE: 0.381)

Spam vs Ham Message Classification w/Custom FNet (F1 Score: 0.92)

Sentiment Analysis w/CatBoostClassifier (F1 Score: 0.703)

Complaint Analysis w/Ensemble Model (CatBoost, LR) (F1 Score: 0.86)

News Analysis w/Tensorflow (DistilBERT fine-tuning) (F1 Score: 0.89)

Emotion Classification w/LogisticRegression

Spam Mail Detection w/Tensorflow (DistilBERT fine-tuning) (F1 Score: 0.92)

2.5 Other Deep Learning Projects (Anomaly Detection, Image Captioning, Multiple Instance Learning etc.)

PaliGemma 3B for Image Captioning

Ford Motor Data Anomaly Detection with AutoEncoder

Satellite Image Captioning (ViT, Bi-LSTMs)

Molecule Activity, Deep Multiple Instance Learning

Cloud Classification (Involution Neural Network)

Car Detect w/Deep Multiple Instance Learning

Happy Detection w/Deep Multiple Instance Learning

---

3. Hybrid Model (Deep Learning and Machine Learning) Projects

(QSAR) Renin Activity (ChEMBL286) Classification w/Ensemble Model(CNN + CatBoost) (F1-Score: 0.84)

Leaf Disease Detection w/Hybrid Model (ViT, PCA, SVM) (F1 Score: 0.92)

Flower Detection w/Hybrid Model (ViT, CatBoost, SHAP) (F1 Score: 0.96)

Skin Cancer Detection w/Hybrid Model (ConvMixer, CatBoost, SHAP)

Smoking Detection w/Hybrid Model (ViT, XGBoost, SHAP) (F1 Score: 0.96)

Disease Severity Hybrid Classifier (ViT,CatBoost) (F1 Score: 0.75)

Diamond Detect w/Hybrid Model (ViT,CatBoost,SHAP) (F1 Score: 0.97)

Mammals Classification w/Ensemble Deep Learning (F1 Score: 0.92)

---

4. Online/Incremental Learning Projects

Cryptocurrency (AVAX) price prediction with Incremental/Online Learning

Smoking Image Detection w/Online Learning (River) (F1 Score: 0.95)

Turbine Power Output Forecasting w/Online Learning (River)

Tesla Stock Price Prediction w/Online Learning

---

5. Machine Learning Theory

Simple New Sample Generation from MNIST w/KDE

Simple New Sample Generation FashionMNIST w/KDE

Minkowski vs Hassanat Distance Metric Implementation w/KNN

Proof of the Entropy of The Gaussian Distribution Implementation

Basic Decision Tree Project and the ccp_alpha parameter tuning (Coursera Project Network)

---

6. Algorithmic Trading

Getting Binance Current Coins Prices Volumes

---

Competitions

	My degree	Type of Competition	My solution algorithm	Link
Forecasting Mini-Course Sales	253/1172	Time Series Forecasting	Deep Learning	https://www.kaggle.com/competitions/playground-series-s3e19
Predict CO2 Emissions in Rwanda	256/1440	Regression	CatBoost, LightGBM	https://www.kaggle.com/competitions/playground-series-s3e20
Petals to the Metal - Flower Classification on TPU	40/118	Classification	Ensemble Deep Learning	https://www.kaggle.com/competitions/tpu-getting-started