TensorFlow-Object-Detection-API-Tutorial-Train-Multiple-Objects-Windows-10
TensorFlow-Object-Detection-API-Tutorial-Train-Multiple-Objects-Windows-10 copied to clipboard
Issue in generating tfrecord
Traceback (most recent call last):
File "generate_tfrecord.py", line 102, in
SORRY FOR ASKING THIS DOUBT IN THE COMMENT, I POSTED IT VOID BY MISTAKE
I've had this before. Have you ever copied your xml files to make it easier to label data for similiar images? Make sure you check all the xml files to see they have the correct name and point to the right file because if you copy xml files and edit the annotations the file names don't update sometimes.
I've had this before. Have you ever copied your xml files to make it easier to label data for similiar images? Make sure you check all the xml files to see they have the correct name and point to the right file because if you copy xml files and edit the annotations the file names don't update sometimes.
I eventually got everything right, the tfrecord the training etc, and I will be honest, i dont know what I did. I think I deleted all the xml files, train labels and test labels and re-labeled the images. The problem is solved but can you explain what exactly was its cause? I wanna know since i have no idea how it got rectified just by relabeling
It is more than likely that your XML files had the wrong names in them so they were pointing to files that either didn't exist or were in the wrong folder.
I have created dataset for one category with 2500 images.I have divided the images into train(80% = 2000 ) and test (20% = 500) folder. CSV is generated successfully but tf record for test is generated successfully.When i want to generate tf record for train is not generating successfully.
But when,
I have created dataset for one category with 1000 images.I have divided the images into train(80% = 800 ) and test (20% = 200) folder CSV is generated successfully.CSV and TF record generated successfully for both train and test.
I dont know why it is happening. Can any one help me out with this?
@SouradipBh Hey,
long time ago, but just for interest, did you find out why? I mean is it necesarry how the images are named for exmaple image1.jpg or image_1.jpg?