Interpretable multi-label eye disease diagnosis on ODIR-5K: prototype learning from scratch vs ResNet-18 + CBAM, with localisation maps and text reports.

AI-powered weed detection using YOLO26 and the MH-Weed16 dataset, with computer vision, agronomic feature engineering, model optimization, and edge deployment for precision agriculture.

A Japanese character recognition system built from scratch using a Convolutional Neural Network (CNN). The current version focuses on recognizing Hiragana characters from user-drawn input, with the goal of eventually expanding the system to Katakana, Kanji, and complete Japanese word recognition.

Blind modulation classification on raw IQ data: 1D CNN vs higher-order cumulants across −10…20 dB SNR, with experiments on observation length and domain shift (multipath, phase noise, IQ imbalance, AdaBN). PyTorch + NumPy, from scratch.

Modular PyTorch multi-layer classifier built from scratch to study training dynamics, hyperparameter tuning, and troubleshooting (overfitting, underfitting, and batch instability) on the Iris dataset

A curated list of free AI tools and APIs — no credit card required

Fisher information geometry and Gauss–Bonnet verification on VAE latent spaces (Bachelor’s thesis)

Posterior sampling with convolutional neural network-based plug-and-play regularization with applications to poststack seismic inversion

High-Performance CUDA C++ GPU kernel library for fused FlashAttention-2, GEMM, and FP8 tensor operations.

A hands-on LLM engineering guide covering Transformer internals, building a small GPT from scratch with PyTorch, prompting, fine-tuning concepts, RAG, AI agents, evaluation, safety, Mamba, and multimodal demos — all in an interactive Jupyter notebook.

Scalable Graph Neural Network (GNN) training pipeline with PyTorch Geometric and dynamic neighborhood sampling.

从线性代数到大模型：用几何直觉讲透向量、矩阵、SVD，再一路走到 Attention、LoRA、KV Cache 与量化。含 81 篇概念文档、358 道可判分题的 Astro 学习站。

Unofficial, modern, dependency-clean toolkit for BDD100K dataset.

From a single perceptron to tool-using agents, in 27 notebooks, building every piece from scratch.

PyTorch deep learning notebooks applied to finance: tensors and training loops, classification, computer vision, and a CNN that classifies candlestick chart patterns.

NeuroScan AI: an end-to-end deep learning web app that classifies brain MRI scans into Glioma, Meningioma, Pituitary tumor, or No Tumor. Built with a fine-tuned VGG16 CNN (transfer learning) in TensorFlow/Keras and served through Flask, with confidence scores, class probabilities and a dark dashboard UI. Educational project, not for diagnosis.

🧠 The Complete, Production-Grade LLM Encyclopedia: Transformer Architectures, 3D Parallelism, Post-Training RL (GRPO/PPO), Synthetic Data, MoE, and 30+ Landmark Research Papers.

Deterministic, sub-microsecond bare-metal neural compute engine in #![no_std] Rust booting directly from UEFI.

IPython - Neural Gesture Analysis with Webcam-based User-input Recognition.

Prédiction du salaire annuel d’un développeur à partir de ses compétences et caractéristiques professionnelles (Stack Overflow 2018) : nettoyage, EDA, feature engineering, comparaison de modèles ML (Régression Linéaire, Random Forest, XGBoost, TensorFlow), sélection du meilleur modèle, API REST Flask et interface React interactive.

Hybrid intrusion detection system combining Random Forest and CNN-LSTM models with a Flask dashboard and forensic logging.

Proyek ini merupakan sistem pemantauan dan penghitungan kendaraan (motor) parkir otomatis berbasis YOLOv8 dan metode Line Crossing yang dilengkapi dengan fitur deteksi cahaya rendah.

InfraPulse — A unified multimodal AI framework for predictive health monitoring of railway tracks, bridges, and dams using deep learning, anomaly detection, risk assessment, and intelligent infrastructure analytics.

Русские переводы открытых книг по воплощённому ИИ и робототехнике: Embodied AI (Datawhale), Dive into Deep Learning, Physical AI Systems. Онлайн-версия на mdBook.

Top Deep Learning Frameworks (Opensource) Star if you like it!

Multimodal clinical decision support platform: image, symptom and voice intake, two-stage MobileNet + ViT pipeline, and consent gated LangGraph agents for patients and doctors.

A curated catalog of AI learning resources — courses and the papers, books and docs that go with them — grouped by skill.

MCP server for AI mammogram triage: prioritizes BI-RADS 4-5 cases so women reach a radiologist in time. Research tool, not a medical device.

Deep Learning Based Low Light Image & Video Enhancement

An end-to-end deep learning framework for multi-class classification and pathology localization of dental OPG radiographs using transfer learning and Grad-CAM.