Thank You Again! I will experiment with this particular OCR to see if it can handle difficult 'images' such a PDF page scans / images !
@Sara-s-playground2 ай бұрын
Thank you so much for trying this out and leaving comments afterwards! They really help!
@davidtindell9502 ай бұрын
@@Sara-s-playground YWSM ! 'YourWelcomeSoMuch' Also, EasyOCR did handle my test PDF 'image' pages fairly well !
@davidtindell9502 ай бұрын
ByeTheWay (BTW/BtW): If your channel offered a "Buy a Coffee" ('ko-fi') link then we could show our appreciation for your efforts and 'Support the Channel'. Whether you drink coffee or tea, or both, You could also use the accumulated funds to buy external SSD's, memory upgrades, NVidia GPU's (... several month's worth of coffees ...) etc.
@Sara-s-playground2 ай бұрын
Thank you so much for giving me this idea. To be honest, I am planning to implement it, as AI projects always do require resources and this is at least GPUs. Soon I will leave the link under my videos. Your comments are already a big appreciation for me.
@Sara-s-playground2 ай бұрын
huggingface.co/spaces/Sara579/ocr link for the hugging face space www.jaided.ai/easyocr/ Documentation and Demo of EasyOCR
@davidtindell9502 ай бұрын
P.S. I re-installed 'PyTorch' + CUDA (12.4) in my "Sara" conda env and NOW my local 'EasyOCR' runs "FAST" even when processing pages of "PDF-JPEG images ! Thank You.
@Sara-s-playground2 ай бұрын
Wow, you named conda environment "Sara" )) so you had an issue with it previously? If so, I am glad you could resolve that issue. PyTorch and CUDA incompatibility problem was a headache for me as well, but I could figure out.
@davidtindell9502 ай бұрын
@@Sara-s-playground I name conda envs after the YT Channel, e.g. "Tim" for "Tech-with-Tim", "John" for "JohnnyCode" etc. As you know, keeping conda envs separate is important to isolate and prevent library and lib versions conflicts, include 'Torch' but also ChromaDB, FAISS, etc.
@davidtindell9502 ай бұрын
NOTE: It also ran relatively well under 'streamlit' on my own laptop! Not Surprisingly, I do obtain the WARNING: "Neither CUDA nor MPS are available - defaulting to CPU. Note: This module is much faster with a GPU. "!