Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

about tf_records_generation.py for sentinel-hub example #133

Open
bobleegogogo opened this issue Mar 13, 2019 · 2 comments
Open

about tf_records_generation.py for sentinel-hub example #133

bobleegogogo opened this issue Mar 13, 2019 · 2 comments

Comments

@bobleegogogo
Copy link

Hello I am looking at the Preparing Data with Sentinel Hub example, since that label-maker also support to prepare the mask annotation of buildings as shown in this examples, do you have some suggestion how to conver the images and lables into tensorflow record for fine-tuning the pre-trained model ?

I have seen some similar python script such as models/research/object_detection/dataset_tools/create_pet_tf_record.py, while to modify this based on label-maker results might be missing :)

@drewbo
Copy link
Contributor

drewbo commented Mar 13, 2019

@bobleegogogo have you tried running the tf_records_generation.py script with the Sentinel Hub data? I think the major change you would need to make from the Sentinel Hub example is to us ml_type as object_detection but otherwise this should be similar to the other building detection example.

@bobleegogogo
Copy link
Author

@drewbo Thanks for your reply. Actually, I was trying to use label-maker to prepare training samples for Mask-RCNN, this is why I need to use ml_type as segmentation. Then the tf_records_generation.py cannot meet the demand, since the training samples should contain mask and bounding boxes, not only the bounding boxes. I found some similar script here: create_mask_rcnn_tf_record.py

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants