-
Notifications
You must be signed in to change notification settings - Fork 87
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
Issue with accuracy and loss during training #17
Comments
I also encountered this problem, how many samples do you have for your training set? |
I used around around 500 images on my training set. How much did you use ? |
around 10 images,Do you have GPU training? |
Yes, I tried training it on a GPU. I tried making some changes to some of the parameters as well. Nothing worked for me |
I'm trying to run with GPU now, but I have some problems, can you teach me how to run with GPU? |
Yes, what is the issue you are facing ? |
First of all, if you use GPU training, do you need to change some code in the train.py? |
Not Needed, just make sure CUDA is set up properly and that tensorflow is detecting your GPU. After that just make sure the model weights file is detected and it will work. |
Yes, I feel CUDA environment is not configured well. |
Can you give me your dataset? Maybe I'll give it a try. |
Is the accuracy rate.improve now? |
I seemed to have found a solution to improve the accuracy,Could you share your dataset? i can try the effect for you |
I used the ICDAR 2013 dataset, https://www.tamirhassan.com/html/dataset.html |
This code needs json format, how to convert ICDAR2013 files into json format? How is your method done, can you give me a copy of the converted data or provide some help? |
Hello,
I tried training a model with my own data after annotating it using label-me. But while training I observed that my training loss is not converging nor is my accuracy increasing. It is stuck within a range of values (0.45 - 0.55 for accuracy). Any Idea why this is happening ?
The text was updated successfully, but these errors were encountered: