The data set readout is abnormal when running Faster-RCNN in TLT

Hi,when I use Faster-RCNN to train my own dataset in TLT, I get the following error:

any idea?thanks

Can you attach the full log? I cannot see any error in your screenshot file.

the error is “Corrupt JPEG data: 608 extraneous bytes before marker 0xfe”
only a very small part of my data set is used for training

Seems that there is something wrong in your dataset. You can try other public dataset.
You can also narrow down or figure out via searching “608 extraneous bytes before marker” in google.

thanks,i tried to figure out “608 extraneous bytes before marker” in google, but the relevant information was scarce, so I decided to change the data set.