I still remember the thrill of submitting my first deep learning paper. It was a journey that lasted almost two months, and I learned a lot along the way. In this post, I want to share my experience with you, and hopefully, it will help you if you’re an undergraduate student like me, venturing into the world of deep learning.
Learning Quickly
The first step was to learn the relative knowledge quickly. I followed my boss’s advice and read about 100 papers related to my project, which was enhancing fundus images using deep learning methods. It was a time-consuming process, but it helped me understand the importance of research papers.
Choosing the Right Method
Next, I had to choose the main method for my project. I noticed that diffusion models, GANs, and transformers were frequently used in papers, so I decided to learn them quickly through YouTube tutorials. I also read typical papers about these methods to understand their core knowledge.
Finding Innovation
After that, I had to find the core innovation for my paper. I summarized the shortcomings of existing papers on enhancing fundus images with diffusion models and identified areas for improvement. There are three ways to create a paper: proposing a new method, fixing shortcomings in existing methods, or fusing different methods.
Writing Code
The fourth step was to write code. I quickly went through the PyTorch tutorial and understood the basics of coding. Then, I used Gemini to write the code for my diffusion model.
Running Comparisons
Next, I had to run comparison codes to show that my method was better than existing ones. I used typical methods like Pix2PixGAN and Stable Diffusion and adapted them to my dataset.
Training and Writing the Paper
Finally, I trained my model and wrote the paper. My boss helped me with the writing process, and I was responsible for creating the figures. I used PPT to make the figures, which was a good and easy way to do it.
Conclusion
Submitting my first deep learning paper was a wonderful experience. I learned a lot, and it helped me to understand the importance of research papers, choosing the right method, finding innovation, writing code, and running comparisons. If you’re an undergraduate student like me, I hope this post will inspire you to venture into the world of deep learning.
*Further reading: What is Deep Learning?*