Awesome Deep Learning Projects

The Gallery by Weights & Biases features curated machine learning reports by researchers exploring deep learning techniques, Kagglers showcasing winning models, and industry leaders sharing best practices.

💥 This hacktoberfest we are open for contributions. Check out our CONTRIBUTING.md to learn more. 💥

awesome-dl-projects by wandb is a collection of the code that accompanies the reports.

Report	Description	Author
Survival of the Fittest CNN Model	Optimize the hyperparameters of a CNN with Variable Length Genetic Algorithm.	Aritra Roy Gosthipaty
Under the hood of RNNs	A NumPy implementation of RNN. Gradients and connectivity has been visualised too.	Aritra Roy Gosthipaty
Under the hood of LSTMs	A NumPy implementation of LSTMs. Difference of architectures and the pros of LSTMs have been discussed.	Aritra Roy Gosthipaty
Get Started with TensorFlow Lite Examples Using Android Studio	A step-by-step guide towards running example apps on your phone using Android Studio, TensorFlow Lite, and USB debugging.	Ivan Goncharov
Part 2: Deep Representations, a way towards neural style transfer	A top down approach to conceiving neural style transfer.	Aritra Roy Gosthipaty and Devjyoti Chakraborty
Part 1: Deep Representations, a way towards neural style transfer	A top down approach to conceiving neural style transfer.	Aritra Roy Gosthipaty and Devjyoti Chakraborty
Hyperparameter Optimization for Huggingface Transformers	The report explains three strategies for hyperparameter optimization for Huggingface Transformers.	Ayush Chaurasia
Ray Tune: Distributed Hyperparameter Optimization at Scale	How to use Ray Tune with W&B to run an effective distributed hyperparameter optimization pipeline at scale.	Ayush Chaurasia and Lavanya Shukla
Modern Scalable Hyperparameter Tuning Methods	A comparison of Random search, Bayesian search using HyperOpt, Bayesian search combined with Asynchronous Hyperband and Population Based Training.	Ayush Chaurasia
Train HuggingFace Models Twice As Fast	This reports summarizes our 14 experiments + 5 reproducibility experiments regarding 2+1 optimizations to reduce training time.	Michaël Benesty
Distilling Knowledge in Neural Networks	This report discusses the compelling model optimization technique - knowledge distillation with code walkthroughs in TensorFlow.	Sayak Paul
Using SimpleTransformers for Common NLP Applications	Explore Language Modeling, Named Entity Recognition, Question Answering with the SimpleTransformer library.	Ayush Chaurasia
SimpleTransformers: Transformers Made Easy	Simple Transformers removes complexity and lets you get down to what matters – model training and experimenting with the Transformer model architectures.	Ayush Thakur
Sentence Classification With Huggingface BERT and W&B	In this tutorial, we’ll build a near state of the art sentence classifier leveraging the power of recent breakthroughs in the field of Natural Language Processing.	Ayush Chaurasia
3D Image Inpainting	A novel way to convert a single RGB-D image into a 3D image.	Ayush Thakur
Unsupervised Visual Representation Learning with SwAV	In this report, we explore the SwAV framework, as presented in the paper "Unsupervised Learning of Visual Features by Contrasting Cluster Assignments" by Caron et al. SwAV is currently the SoTA in self-supervised learning for visual recognition. We also address common problems in existing self-supervised methods.	Ayush Thakur and Sayak Paul
DeepFaceDrawing: An Overview	This is an overview report to break down the key components of DeepFaceDrawing	Ayush Thakur
Tips for Kagglers From Winning a Solo Silver Medal	How I got started with the competition, my struggles, journey to the final solution and everything I learned in between.	Aman Dalmia
Part 2 – Comparing Message Passing Based GNN Architectures	In this report, we shall review various message-passing based GNN architectures and compare them using Sweeps by Weights and Biases.	Yash Kotadia
Part 1 – Introduction to Graph Neural Networks with GatedGCN	This report summarizes the need for Graph Neural Networks and analyzes one particular architecture – the Gated Graph Convolutional Network.	Yash Kotadia
An Introduction to Adversarial Examples in Deep Learning	This report provides an intuitive introduction to adversarial examples, discusses a wide variety of different adversarial attacks and, most notably, provides advice on defending against them.	Sayak Paul
Text Recognition with CRNN-CTC Network	This report explains how to detect & recognize text from images.	Rajesh Shreedhar Bhat
Visualize Scikit Models	Visualize your scikit-learn model's performance with just a few lines of code	Lavanya Shukla
DeOldify	Understand how the DeOldify model works, reproduce and visualize the results obtained by the author.	Boris Dayma
HuggingTweets - Train a model to generate tweets	In this project, we fine-tune a pre-trained transformer on tweets using HuggingFace Transformers – a popular library with pre-trained architectures and frameworks for NLP.	Boris Dayma
Organize Your Machine Learning Pipelines with Artifacts	In this report, we will show you how to use W&B Artifacts to store and keep track of datasets, models, and evaluation results across machine learning pipelines.	Ayush Chaurasia
Weights & Biases and Ray Tune: Distributed hyperparameter optimization at scale	How to use Ray Tune with W&B to run an effective distributed hyperparameter optimization pipeline at scale.	Ayush Chaurasia and Lavanya Shukla
Towards Representation Learning for an Image Retrieval Task	This report explains self-supervised and regularized supervised image retrieval with the help of the latent space of an autoencoder.	Aritra Roy Gosthipaty and Souradip Chakraborty
Build the World's Open Hedge Fund by Modeling the Stock Market	In this report, we show you how to get started with Numerai, a crowdsourced AI hedge fund and compete on the hardest data science tournament on the planet using Weights & Biases.	Carlo Lepelaars
Understanding the Effectivity of Ensembles in Deep Learning	The report explores the ideas presented in Deep Ensembles: A Loss Landscape Perspective by Stanislav Fort, Huiyi Hu, and Balaji Lakshminarayanan.	Ayush Thakur and Sayak Paul
Plunging into Model Pruning in Deep Learning	This report discusses pruning techniques in the context of deep learning.	Sayak Paul
Visualizing Confusion Matrices with W&B	Using Keras with Weights & Biases, plot a confusion matrix at every step of model training and see where your algorithm is wrong.	Mathïs Fédérico
Experiments with OpenAI Jukebox	Exploring generative models that create music based on raw audio.	Ishaan Malhi
The Power of Random Features of a CNN	This report presents a number of experiments based on the ideas shown in https://arxiv.org/abs/2003.00152 by Frankle et al.	Sayak Paul
The Al-Dente Neural Network: Part I	Much like making pasta, training a neural network is easy to learn but takes a lifetime to master. What follows is probably the best recipe to make your own Al-Dente Neural Net, courtesy of Andrej Karpathy.	Sairam Sundaresan
A Comparative Study of Activation Functions	Walking through different activation functions and comparing their performance.	Sweta Shaw
Generating Digital Painting Lighting Effects via RGB-space Geometry	Exploring the paper "Generating Digital Painting Lighting Effects via RGB-space Geometry" in which the authors propose an image processing algorithm to generate digital painting lighting effects from a single image.	Ayush Thakur
Two Shots to Green Screen: Collage with Deep Learning	Train a deep net to extract foreground and background in natural images and videos	Stacey Svetlichnaya
A Step by Step Guide to Tracking Hugging Face Model Performance	A quick tutorial for training NLP models with HuggingFace and & visualizing their performance with Weights & Biases	Jack Morris
A Tale of Model Quantization in TF Lite	Model optimization strategies and quantization techniques to help deploy machine learning models in resource constrained environments.	Sayak Paul
Drought Watch Benchmark Progress	Developing the baseline and exploring submissions to the Drought Watch benchmark	Stacey Svetlichnaya
Who is Them? Text Disambiguation with Transformers	Using HuggingFace to explore models for natural language understanding	Stacey Svetlichnaya
Lightning Kitti	Semantic segmentation on Kitti dataset with Pytorch-Lightning	Boris Dayma
Interpretability in Deep Learning with W&B - CAM and GradCAM	This report will review how Grad-CAM counters the common criticism that neural networks are not interpretable. We'll review feature visualization, class activation maps and implement a custom callback that you can use in your own projects.	Ayush Thakur
Adversarial Policies in Multi-Agent Settings	One way to win is not to play the game	Stacey Svetlichnaya
Bounding Boxes for Object Detection	How to log and explore bounding boxes	Stacey Svetlichnaya
Deep Q Networks with the Cartpole Environment	A brief explanation of the DQN algorithm for reinforcement learning, focusing on the Cartpole-v1 environment from OpenAI gym.	Jari
Using simpleTransformer on common NLP applications	Explore Language Modeling, Named Entity Recognition, Question Answering with distilbert from the simpleTransformer library.	Ayush Chaurasia
Transfer Learning with EfficientNet family of models	Learn to use the EfficientNet family of models for transfer learning in TensorFlow using TFHub.	Sayak Paul
Automate Kaggle model training with Skorch and W&B	Skorch combines the simplicity of scikit, with the power of pytorch and makes for a great framework to use in Kaggle competitions	Ayush Chaurasia
EvoNorm layers in TensorFlow 2	Experimental summary of my implementation of EvoNorm layers proposed in https://arxiv.org/pdf/2004.02967.pdf.	Sayak Paul
When Inception-ResNet-V2 is too slow	Some versions of Inception parallelize better than others	Stacey Svetlichnaya
Using W&B in a Kaggle Competition	In this tutorial, we’ll see how you can use W&B in a Kaggle competition. We'll also see how W&B's Scikit-learn integration enables you to visualize performance metrics for your model with a single line of code. Finally, we'll run a hyperparameter sweep to pick the best model.	Ayush Chaurasia
Image Masks for Semantic Segmentation	How to log and explore semantic segmentation masks	Stacey Svetlichnaya
COVID-19 research using PyTorch and W&B	How to train T5 on SQUAD with Transformers and Nlp	Ayush Chaurasia
Video to 3D: Depth Perception for Self-Driving Cars	Unsupervised learning of depth perception from dashboard cameras	Stacey Svetlichnaya
The View from the Driver's Seat	Semantic segmentation for scene parsing on Berkeley Deep Drive 100K	Stacey Svetlichnaya
Meaning and Noise in Hyperparameter Search	How do we distinguish signal from pareidolia (imaginary patterns)?	Stacey Svetlichnaya
Kaggle Starter Kernel - Jigsaw Multilingual Toxic Comment Classification	This report presents a comparison between three models, trained to compete on Kaggle's Jigsaw Multilingual Toxic Comment Classification.	Sayak Paul
Sentence classification with Huggingface BERT and W&B	How to train T5 on SQUAD with Transformers and Nlp	Ayush Chaurasia
Visualizing and Debugging Neural Networks with PyTorch and W&B	In this post, we’ll see what makes a neural network underperform and ways we can debug this by visualizing the gradients and other parameters associated with model training. We’ll also discuss the problem of vanishing and exploding gradients and methods to overcome them.	Ayush Thakur
Track Model Performance	In this report, I'll show you show you can visualize any model's performance with Weights & Biases. We'll see how to log metrics from vanilla for loops, boosting models (xgboost & lightgbm), sklearn and neural networks.	Lavanya Shukla
Log ROC, PR curves and Confusion Matrices with W&B	You can now log precision recall and ROC curves, and confusion matrices natively using Weights & Biases. You can also use our heatmaps to create attention maps.	Lavanya Shukla
Visualize models in TensorBoard with Weights and Biases	In this article, we are going see how to spin up and host a TensorBoard instance online with Weights and Biases. We'll end with visualizing a confusion matrix in TensorBoard.	Lavanya Shukla
Visualizing and Debugging Neural Networks with PyTorch and W&B	In this post I'll show you how to use wandb.Molecule, to visualize molecular data with Weights and Biases.	Nicholas Bardy
Visualize Model Predictions	In this report, I'll show you how to visualize a model's predictions with Weights & Biases – images, videos, audio, tables, HTML, metrics, plots, 3d objects and point clouds.	Lavanya Shukla
Use Pytorch Lightning with Weights & Biases	PyTorch Lightning lets you decouple the science code from engineering code. Try this quick tutorial to visualize Lightning models and optimize hyperparameters with an easy Weights & Biases integration.	Ayush Chaurasia
Evaluating the Impact of Sequence Convolutions and Embeddings on Protein Structure Prediction	A vignette on recent work using deep learning for protein structure prediction. March 26, 2020.	Jonathan King
Towards Deep Generative Modeling with W&B	In this report, we will learn about the evolution of generative modeling. We'll start with Autoencoders and Variational Autoencoders and then dive into Generative Adversarial Modeling.	Ayush Thakur
Exploring ResNets With W&B	Post by Ayush Chaurasia	Lavanya Shukla
NeRF – Representing Scenes as Neural Radiance Fields for View Synthesis	In the Representing Scenes as Neural Radiance Fields for View Synthesis paper, the authors present a method that achieves state-of-the-art results for synthesizing novel views of complex scenes by optimizing an underlying continuous volumetric scene function using a sparse set of input views.	Lavanya Shukla
How Efficient is EfficientNet?	Evaluating the EfficientNet Family on a Smaller ImageNet-like Dataset	Ajay Arasanipalai
Distributed training in tf.keras with W&B	Explore the ways to distribute your training workloads with minimal code changes and analyze system metrics with Weights and Biases.	Sayak Paul
Reproducible Models with W&B	Discover simple techniques to make your ML experiments as reproducible as possible.	Sayak Paul
Effects of Weight Initialization on Neural Networks	In this article, we’ll review and compare a plethora of weight initialization methods for neural nets. We will also outline a simple recipe for initializing the weights in a neural net.	Sayak Paul
An Introduction to Image Inpainting using Deep Learning	In this report, we are going to learn how to do “image inpainting”, i.e. fill in missing parts of images precisely using deep learning.	Ayush Thakur and Sayak Paul
Distributed Training	Getting started with distributed training in Keras	Stacey Svetlichnaya
Curriculum Learning in Nature	Applying human learning strategies to neural nets on iNaturalist 2017	Stacey Svetlichnaya
Fashion MNIST	Explore various hyperparameters of a CNN trained on Fashion MNIST to identify 10 types of clothing	Stacey Svetlichnaya
Classify the Natural World	Training and fine-tuning convolutional networks to identify species beyond ImageNet	Stacey Svetlichnaya
Colorizing Black and White Images	How can we add realistic color to black & white images? Explore the effect of up-convolutions, weight decay, and deeper architectures.	Nicholas Bardy
Text Generation with Sequence Models	Explore network architectures and training settings for character-based text generation. Compare RNNs, GRUs, and LSTMS, with different depths, layer widths, and dropout. Also consider the training data length, sequence length, and number of sequences per batch.	Carey
RNNs for Video Understanding	Comparing various recurrent models in Pytorch on YouTube videos	rchavezj
FastText Nanobot for the Transformer Age	Integrate FastText with W&B to visualize incredibly efficient natural language processing	Stacey Svetlichnaya
Dropout in PyTorch – An Example	Regularize your PyTorch model with Dropout	Ayush Thakur
Compare & monitor fastai2 models	Exploring generative models that create music based on raw audio.	Boris Dayma
Generate Meaningful Captions for Images with Attention Models	Image captioning has many use cases that include generating captions for Google image search and live video surveillance as well as helping visually impaired people to get information about their surroundings.	Rajesh Shreedhar Bhat and Souradip Chakraborty
Train HuggingFace models twice as fast	This reports summarizes our 14 experiments + 5 reproducibility experiments regarding 2+1 optimizations to reduce training time.	Michaël Benesty
Build the world's open hedge fund by modeling the stock market.	In this report, we show you how to get started with Numerai, a crowdsourced AI hedge fund and compete on the hardest data science tournament on the planet using Weights & Biases.	Carlo Lepelaars
Use GPUs with Keras	A short tutorial on using GPUs for your deep learning models with Keras	Ayush Thakur
Implementing and tracking the performance of a CNN in Pytorch - An Example	A guide to implementing and tracking the performance of a Convolutional Neural Network in Pytorch.	Ayush Thakur
Measuring Mode Collapse in GANs	Evaluate and quantitatively measure the GAN failure case of mode collapse - when the model fails to generate diverse enough outputs.	Kevin Shen