paddlepaddle/awesome-deeplearning

This repository is a one-stop learning hub for deep learning, created by the team behind PaddlePaddle, Baidu's open-source AI framework. The content is written primarily in Chinese and covers everything from beginner introductions to advanced industry applications. It is organized around four main categories: courses, books, a knowledge encyclopedia, and practical case studies. The courses section includes a free online video course called "Zero-Foundation Practical Deep Learning" with around 20 hours of recorded lessons, taught by Baidu engineers and researchers. There is also a companion printed textbook published by Tsinghua University Press. A separate series of featured courses covers Transformer-based models in depth, walking through architectures like BERT, GPT, ELMo, RoBERTa, and several vision Transformers such as ViT and Swin Transformer. The knowledge section, called "Deep Learning 100 Questions," is a Q&A reference covering fundamentals, advanced topics, applications, and reinforcement learning concepts. It also includes an interview preparation guide. All of this material is hosted on the Paddlepedia documentation platform. The practical examples section is called the Paddle Industry Practice Sample Library. It contains end-to-end project notebooks covering real-world applications in three areas: smart city (fire and smoke detection, safety helmet detection), smart manufacturing (steel defect segmentation, robotic grasping), and internet (financial report recognition and keyword extraction). All notebooks are runnable online through Baidu's AI Studio platform and are kept updated to match the latest version of PaddlePaddle. The repository also includes a PaddlePaddle adaptation of the well-known open textbook "Dive into Deep Learning," converting its original code examples from a different framework to PaddlePaddle. This adaptation requires only basic math and Python knowledge to follow along. The full README is longer than what was shown.