Deep Learning Projects with Source Code

Deep Learning Architectures in Our Collection

Each project focuses on a specific architecture and problem domain. You get to see how the architecture works in practice, not just in theory.

Convolutional Neural Networks (CNNs)

CNN projects handle image classification (medical images, plant diseases, face recognition), object detection (YOLO-based), and image segmentation. Models use architectures like ResNet, VGG, EfficientNet, and custom CNNs. Transfer learning projects show you how to fine-tune pre-trained models on your own dataset with just a few hundred images.

Transformers and Attention Models

Transformer projects cover text classification, question answering, summarization, and named entity recognition. Some use pre-trained models from Hugging Face (BERT, GPT-2, T5) with fine-tuning scripts. Others build transformer architectures from scratch so you understand the attention mechanism at the code level.

Generative Models (GANs and VAEs)

GAN projects include face generation with DCGAN, style transfer with CycleGAN, image super-resolution with SRGAN, and text-to-image generation. Each project includes the generator and discriminator architectures, training loop with loss visualization, and a script for generating new samples from the trained model.

Sequence Models (LSTMs and GRUs)

Sequence model projects predict stock prices, generate text, compose music, and classify time series data. You'll see proper sequence padding, embedding layers, bidirectional LSTM configurations, and attention mechanisms added to recurrent networks.

Training Deep Learning Models

All projects include pre-trained weights, so you don't need to train from scratch. But if you want to retrain — maybe with a different dataset or modified architecture — training scripts are included. For GPU-intensive training, we provide Google Colab notebooks with GPU runtime pre-configured.

Hardware Requirements

Inference (running predictions) works on any modern laptop's CPU. Training CNNs and Transformers benefits from a GPU, which is why we include Colab notebooks. Projects specify estimated training time on both CPU and GPU so you can plan accordingly.

Model Evaluation

Each project includes evaluation scripts that generate accuracy/loss curves, confusion matrices, classification reports, and example predictions. These outputs are formatted for direct inclusion in project reports and presentations.