training dataset
Hello Sarfraz,
Do you have the original dataset that was used to train the classifier?
I wanted to fine-tune this model to add more document types to it. ex: different types of w2 forms.
Looks like the model fails to classify some types of W2 forms.
Hi Syed,
For w2 forms I used original forms from IRS web site with some augmentations. I did not use synthetic data generation, which could have caused the low accuracy. Synthetic data generation improves model accuracy and is needed to make the model better.
Given below are examples
Regards
Got it! Thanks for the response.
May I know how many samples per each class you have used for the classification task?
Is it possible to save this model to onnx format for inferencing?
303 samples for w2 form.
840 samples for 1040
Regards