OCR Form Reader Development with Hailo-8L Chip

As an AI enthusiast, I recently acquired a Raspberry Pi AI Kit with the Hailo-8L chip. I am particularly interested in developing an OCR form(pdf/scanned image) Scanner/reader application using this chip.

I would appreciate it if anyone could confirm on whether this use case is feasible with the Hailo-8L.

Any guidance or resources to assist with this project would be immensely helpful.


Hi @alok.ranjan, I think that the best approach is a two level pipeline. First an object detector (e.g. YOLO), to identify the elements that you want (barcodes, QRcode, etc.), then crop these, and apply OCR on the cropped items.
For OCR we have the lprnet, that is basically an 0-9 digit OCR network. It’s called like that since originally it was designed as the last state of Automated License Plate Reader (ALPR) pipeline.

Hi @Nadav , thank you for your response. I am interested in using Tensorflow ResNet models for text detection and Tensorflow
CRNN_VGG models for text recognition on the Hailo-8L chip. Do you think this is achievable?

Hi @alok.ranjan, any luck in making progress on this project?

Hi @Nadav, I haven’t made much progress yet, but I will keep this thread updated on any significant developments with this project.