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)
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
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)
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)
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)
Getting Binance Current Coins Prices Volumes