search

autonomousvision / occupancy_networks

This repository contains the code for the paper "Occupancy Networks - Learning 3D Reconstruction in Function Space"
https://avg.is.tuebingen.mpg.de/publications/occupancy-networks
MIT License
1.5k stars 292 forks source link

How to use your own data set to train the model？ #6

Closed ChauncyFr closed 4 years ago

ChauncyFr commented 5 years ago

How to use your own data set to train the model？

LMescheder commented 5 years ago

Hi @ChauncyFr , it would be great if you could be a bit more specific. Did you try to follow the instructions in the README? Where did you encounter issues?

ChauncyFr commented 5 years ago

Hi @ChauncyFr , it would be great if you could be a bit more specific. Did you try to follow the instructions in the README? Where did you encounter issues?

First of all, thank you for your reply. I want to use this method for 3D reconstruction of a regular box (for example, a courier box). I also collect photos of the box myself and want to train my own dataset to get the model. I don't know how to pre-process and train my dataset. Can I train my own dataset (migration learning) based on your trained model and then get the model of my own dataset.

I look forward to your answer, thank you again!

LMescheder commented 5 years ago

Take a look at how we preprocess the data in scripts/dataset_shapenet/build.sh (also look at the corresponding section in the readme). When you do this for your dataset, you should be able to apply our code to your data.

ChauncyFr commented 5 years ago

Take a look at how we preprocess the data in scripts/dataset_shapenet/build.sh (also look at the corresponding section in the readme). When you do this for your dataset, you should be able to apply our code to your data. Ok, I still have a question: Can you learn migration based on the model you are training, because my personal data set is relatively small, so I thought about whether I can continue to train my training model. Data sets for better results? For example, mask rcnn provides a pre-training model on the coco dataset. I only need a small amount of data to train my own data based on the pre-trained model, and finally get good results.

LMescheder commented 4 years ago

Closing, because it is duplicate #30.