Projects | AI

Project Overview

Project	Domain	Key Techniques	Data Type
Proj1	Image Classification (Wildlife)	CNN, Binary/Multiclass	Images
Proj2	Transfer Learning (Cassava)	Fine-tuning, Callbacks	Images
Proj3	Object Detection	YOLOv8, Data Augmentation	Images + Annotations
Proj4	Face Recognition	MTCNN, InceptionResNet, Flask	Face Images
Proj5	GANs	Generator/Discriminator	Training Images
Proj6	Diffusion Models	Stable Diffusion, Prompting	Text Prompts

Proj1: Image Classification

Camera Traps

CNN from scratch for wildlife camera trap image classification. Covers binary classification (animal/no animal) and multiclass species identification.

Notebooks

011 - Image as Data: Loading and preprocessing camera trap images
013 - Binary Classification: Animal vs. no animal detection
014 - Multiclass Classification: Species identification

Techniques

CNN from scratch Custom training loops Data preprocessing

Proj2: Transfer Learning

Cassava Disease

Fine-tuning pretrained models for cassava leaf disease detection. Demonstrates transfer learning with callbacks for early stopping.

Notebooks

023 - Multiclass Classification: Disease category prediction
024 - Transfer Learning: Fine-tuning pretrained models
025 - Callbacks: Early stopping and model checkpointing

Models

pretrained_model.pth - Base pretrained weights
model_trained.pth - Fine-tuned model

Techniques

EfficientNet ResNet50 VGG16 Early stopping

Proj3: Object Detection

YOLOv8

Object detection using YOLOv8 with custom training and data augmentation pipelines.

Notebooks

033 - Basic YOLO: Introduction to object detection
034 - Custom Objects: Training on custom datasets
035 - Data Augmentation: Improving model robustness

Techniques

YOLOv8 Bounding boxes Augmentation pipelines

Proj4: Face Recognition

MTCNN + FaceNet

Face detection and recognition with MTCNN and InceptionResNet, deployed as a Flask API.

Notebooks

043 - MTCNN: Face detection
044 - InceptionResNet: Face embedding extraction
045 - Flask API: Web service deployment

Modules

facenet.py - FaceNet embedding model
wq_face_recognition.py - Recognition pipeline
wq_app.py - Flask web application

Techniques

MTCNN InceptionResNet Embeddings Flask API

Proj5: Generative Adversarial Networks

GANs

Training GANs with generator and discriminator networks for image generation.

Notebooks

052 - Vanilla GAN: Basic GAN architecture and training

Models

generator_99.pth - Trained generator weights
discriminator_99.pth - Trained discriminator weights

Techniques

DCGAN Generator Discriminator Adversarial training

Proj6: Diffusion Models

Stable Diffusion

Text-to-image generation using Stable Diffusion with prompt engineering techniques.

Notebooks

062 - Diffusion Pipelines: Text-to-image generation

Techniques

Stable Diffusion Prompt engineering Text-to-image Pipeline configs