about csv&xml based annotations

[nirolada]

I suggested a class of annotations reader for reading and validating annotation files in CSV & XML formats.
What do you think about it? Should I start working on it? Any other suggestions?

[Nelson-Gon]

I think it sounds good. I would only add that we need to figure out how to use these in training but I imagine you've thought about this already. Thanks!

[nirolada]

What do you prefer?

1. Read the data and create a python dictionary for each row.
2. Read the data and create a pandas (a python package like Numpy) dataframe for the whole data.

[Nelson-Gon]

I think a DataFrame would be more resource friendly and possibly easier to work with. That said, I have not yet had time to take a look at the best way to feed this into the Keras training pipeline.

[nirolada]

This is a good point. Can you check and fill me in what type input of annotations does keras accept?

…

On Sun, Mar 7, 2021, 09:50 NelsonGon ***@***.***> wrote: I think a DataFrame would be more resource friendly and possibly easier to work with. That said, I have not yet had time to take a look at the best way to feed this into the Keras training pipeline. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#32 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALPEO5GBT3UK3QZHEC3EHH3TCMV3BANCNFSM4YPLDN6A> .

[Nelson-Gon]

I will look into it more and will respond by Tuesday.

[Nelson-Gon]

@nirolada This is the closest I have found. From the linked StackOverflow question, it seems the best bet is to write a custom data generator. In response to the kind of data, it seems there is no native support for labels with Keras.