deep-anpr
deep-anpr copied to clipboard
matthewearl
Arabic letters unicode issue
Hello everyone I am trying to implement this code for moroccan licence plate detection, the moroccan number plates contain arabic letters and number and i am having trouble generating the dataset of images
UPDATE : I ended up making gen.py work and generated the dataset with the arabic letters using UTF-8 encodings for CHARS but now when i tried launching training i was getting encoding errors (i bypassed them by redefining CHARS as a non encoded string) but now i am getting incompatible tensor shape errors. I am new with Tensorflow can someone help me ?
@ILYASMOUTAWWAKIL
Yes i am, i was able to adapt the code, genrate the dataset and launch the training, i'm waiting on the loss function to converge it might take a while since i don't have gpu.
I posted the code here
And what abt real time detection the code mathew uses is quite hard coding the plate detection .. it takes him a lot of time to process one image while using his GPU .. his solution is not practical for embedded systems ..
On Sun, 24 Mar 2019 17:52 jihadbourassi, [email protected] wrote:
@ILYASMOUTAWWAKIL https://github.com/ILYASMOUTAWWAKIL Yes i am, i was able to adapt the codegenrate the dataset and launch the training, i'm waiting on the loss function to converge it might take a while since i don'thave gpu
i posted the code here https://github.com/jihadbourassi/deep-anpr-for-Moroccan-licence-plates
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/matthewearl/deep-anpr/issues/107#issuecomment-475977349, or mute the thread https://github.com/notifications/unsubscribe-auth/Arihpaamg4jY6E4Wp3UpN_jdya7Ry44Uks5vZ61hgaJpZM4bhAeO .
@ILYASMOUTAWWAKIL Since my academic project is about processing one image at a time and not live detection in a video, I'll be satisfied if I achieve just that. But you're right, I'm aware that Matthew's code is not ideal for real time recognition, I will definitely expand upon it once I am done with my project obligations.
