Training data and training-related errors

[HONG-DEOK]

I am trying to train using sam2.
(I'm trying to train image, not video.)

Environment construction was completed in a Windows environment, and learning data was constructed using tools provided by roboflow.
(reference : https://blog.roboflow.com/fine-tune-sam-2-1/)

However, two errors occur when learning data built through roboflow. Can you please tell me a solution to this?

Error 1 : ERROR 2024-12-24 11:12:28,094 sam2_datasets.py: 63: Caught ValueError in DataLoader worker process 3.

Error 2 : ValueError: Invalid RLE mask representation

Learning is successful using the automobile parts dataset provided by roboflow, but learning is not possible with data built with my own data in roboflow.

I will share one of the json files I created.

img-tif_0_0_jpg_rf_3917c6e1f23f226d655a927307ebb6ca_1.json