Object Detection of alphabets and digits on terrain images and videos by applying Deep learning. Export the output at regular time intervals to a spreadsheet.
We thought about the problem and thought of splitting the problem into 4 parts first processing the video 📹 then proceesing the image 🎞️ then do the alpabhet detection 🧠: then writing to csv.
I had now started the hunt for the dataset and after 2 hours of relentless searching 🔍 we found nothing similar test data 🙁 so we tried something ingenious what we did was take a some photos of the test data then add letters onto it by using a dataset of alhapbets on it by opencv by differning the weight of the sum.
Now here comes the hard part 😨 . I had the dataset already spilt into its section but now I had to load it into an array. Then I augmented the data use image data generator then sent it through the model.
We used our own model so we could hypertune the parameters we had accuracy of about 82 percent 😄.