search

BoiseState-AdaptLab / OCR_4_Forest_Service

Other
2 stars 0 forks source link

Read literature on character recognition #42

Open FloCiaglia opened 3 years ago

FloCiaglia commented 3 years ago

Read this paper

Try Google Cloud

FloCiaglia commented 3 years ago

I have started with the Google Cloud Vision API. The API works very well on some instances but on others, it returns a blank answer. I am going to test if the API works better on the original image or the preprocessed one. Then, I will test on the entire form and compare the results to our pipeline's.

FloCiaglia commented 3 years ago

The results of testing between original and preprocessed images proved that the original yielded better results. So, I on tested three of our forms. The results were better than our pipelines so the plan is to switch the default program to run the Google Cloud Vision API rather than our pipeline.