search

ninadmn / 3D-Segmentation-of-Vertebra-using-Volumetreic-Network

Thesis project on 3D instance segmentation of vertebra using Mask RCNN
8 stars 0 forks source link

Dataset preparation #1

Open JacCubk opened 1 year ago

JacCubk commented 1 year ago

Hello, thank you for your contribution to such a great work. Can you tell me how to prepare a dataset for training?

ninadmn commented 1 year ago

Absolutely, I'd be delighted to assist you. In my scenario, I started off with just the scans, devoid of any ground truth data. I had to generate synthetic ground truth data, specifically masks, by creating 2 KMeans clusters. These clusters focused primarily on differentiating between soft tissue/bone and air. Additionally, I employed image processing techniques such as erosion and dilation, which effectively eliminated small features like pulmonary vessels or noise.

Would you be willing to share your email? I can then provide you with the specific code that helped me get started with this process. Please note that the method I utilised might only be suitable for academic or personal use due to its potential accuracy limitations. However, I hope it will serve as a valuable starting point for your work.

JacCubk commented 1 year ago

Hello!

Thank you very much.

I am a beginner in hyperspectral image processing. I could understand your network, but I do not know how to prepare the dataset, such as the file structure of the dataset. In fact, I plan to collect my own hyperspectral dataset. Therefore, I am eager to learn about your file structure as a reference to prepare my dataset.

Thank you again for your help. This is my email: y.wang.gzu@foxmail.com.

Looking forward to your reply!

Wishing you the best!

JacCubk commented 1 year ago

Thank you again for your help. This is my email: y.wang.gzu@foxmail.com

JacCubk commented 1 year ago

Hello! I have prepared my own dataset, but I don't know how to prepare the data format to match your code. Can you help me?

Thank you again!