ObliqueOCR is a CNN based project which predicts the angle of rotation of the given oblique image and also extracts text from it.
I have fine-tuned pre-trained EfficientNetB7 and MobileNet networks on my a custom dataset provided by Signzy. The Easy-OCR library was used to extract text from the alligned images. Alternatively, I have also used PCA to find the angle of orientation of a given image.