-
Notifications
You must be signed in to change notification settings - Fork 1
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
Out of range: End of sequence #2
Comments
Did you make any changes? That error should already be handled by MonitoredTrainingSession |
no, I don't know why, can you reproduce the result of the paper using the image matting data you send to me, do you finetune the model or just train from scratch |
I am getting around 92% and 85% IOU for my training and validation set, the main difficulty is when it is fed portrait in the wild the results are not convincing, most likely because the lack of variation in the dataset. Currently I am planning to use COCO and/or VOC dataset as well in addition to the portrait dataset. You should probably check out my pix2pix matting repository if interested, it is using u-net encoder-decoder network which has better results in my experiments but it is computationally expensive to fit in client side like mobile phone. |
And I will be training from scratch when I add the new dataset since the data distribution is not exactly the same. |
so for now you finetuned your model using the size 128*128? can you provide your email for convenience |
yes, I was finetuning my model with the same size and sure [email protected] |
thanks, I reproduced your results by your data, keep contact |
Out of range: End of sequence error when come to epoch 55000, I can reproduce the bug
The text was updated successfully, but these errors were encountered: