JaidedAI
EasyOCR is not giving output and exiting before printing anything and also no error on console!!!!
When i run the code it just exiting halfway there not even giving output, please help me
my system config:
- Python 3.12.8
- windows 10
- i5 3gen
- 8gb ram
- 730 NVidia gpu
- 256 SSD
project structure:
input images:
code:
import torch
import easyocr
import cv2
print(f"PyTorch version: {torch.__version__}")
print(f"EasyOCR version: {easyocr.__version__}")
print(f"OpenCV version: {cv2.__version__}")
try:
reader = easyocr.Reader(['en'])
result = reader.readtext('input/d.png')
print(result)
result = reader.readtext('input/image.jpg')
print(f"OCR Result: {result}")
except Exception as e:
print(f"An error occurred: {e}")
output:
PS C:\Users\user\Desktop\my-projects\random-stuff\python-image-translator> python .\testocr.py
PyTorch version: 2.5.1+cpu
EasyOCR version: 1.7.2
OpenCV version: 4.10.0
Neither CUDA nor MPS are available - defaulting to CPU. Note: This module is much faster with a GPU.
PS C:\Users\user\Desktop\my-projects\random-stuff\python-image-translator>
PS C:\Users\user\Desktop\my-projects\random-stuff\python-image-translator> python .\testocr.py
PyTorch version: 2.5.1+cpu
EasyOCR version: 1.7.2
OpenCV version: 4.10.0
Neither CUDA nor MPS are available - defaulting to CPU. Note: This module is much faster with a GPU.
PS C:\Users\user\Desktop\my-projects\random-stuff\python-image-translator>
yes same issue with me, the code works fine on Colab with CPU setting but not on laptop
yes same issue with me, the code works fine on Colab with CPU setting but not on laptop
and the worst part is, not giving any error !!!
then how we solve this issue ?? there is no help on internet
No idea, maybe you could try to reinstall your os.
I'm not sure if it's related to this issue, but in my environment I've encountered a situation where it seems to hang inside reader.readtext. It appears to work with PyTorch 2.2.0, but there have been cases where it doesn't work properly with PyTorch 2.2.1, 2.2.2, and 2.3.0.
with PyTorch 2.2.0, but there have been cases where it doesn't work properly w
please do tell us if you fix the issue.
FYI, in my case, it appeared to hang when calling toTensor in NormalizePAD. https://github.com/JaidedAI/EasyOCR/blob/master/easyocr/recognition.py#L39
However, it doesn't always happen, and since the issue occurred with a certain PIL Image that didn't seem particularly problematic, I still don't really know the cause at this point.
First and foremost, it appears to run without any issues on Linux.
I was testing in an Apple M1 environment, and I observed the following behavior as far as my usage was concerned; The Python version did not seem to have a significant impact, so I will omit it for now (I actually conducted experiments using Python 3.9, 3.11, and 3.12.). Additionally, in my environment, the specific points where the program freezes seemed to vary slightly depending on the input data.
Does it depend on the OS version?
| Ventura (13.5.2) | Sequoia (15.3.1) | |
|---|---|---|
| PyTorch 2.2.0 | OK | OK |
| PyTorch 2.3.0 | NG | NG |
Looking at these results, the OS version does not seem to be a major factor.
Is there a potential issue with the CPU?
Next, I conducted an experiment using Docker to run aarch64 Debian "bullseye" on Sequoia (15.3.1) + M1.
| Debian “bullseye” on Sequoia (15.3.1) + M1 | |
|---|---|
| PyTorch 2.3.0 | OK |
Based on this result, the fact that the CPU is M1 also does not seem to be a determining factor.
While the cause remains unknown, in environments where the issue occurs, one possible workaround could be to use Docker or WSL2 on Windows, or Docker on Mac, to create a Linux environment and run EasyOCR within it.
ok thanks for the info, lets see if i can run in docker, can you provide the docker file
I will include the Dockerfile that I had ChatGPT generate based on what I needed and then manually refined it. I started the docker container, installed additional necessary libraries inside it, and tested an application including EasyOCR. The reason I am using Python 3.11 instead of the latest version is that, in addition to EasyOCR, I am also using an other library that is not yet compatible with Python 3.12.
FROM python:3.11.11-bullseye
ENV DEBIAN_FRONTEND=noninteractive
RUN apt-get update && apt-get install -y \
git \
build-essential
RUN python -m pip install --upgrade pip
WORKDIR /work
RUN python -m venv /venv
ENV PATH="/venv/bin:$PATH"
RUN /venv/bin/pip install -U pip matplotlib jupyter jupyterlab
COPY jupyter_lab_config.py /root/.jupyter/
CMD ["python"]
I came across some interesting information that might be worth trying. Instead of running python something.py, we just run OMP_NUM_THREADS=1 python something.py (If you use jupyter, run OMP_NUM_THREADS=1 jupyter in the same way). This idea comes from https://github.com/microsoft/LightGBM/issues/6595#issuecomment-2315978807 (also https://github.com/pytorch/pytorch/issues/121101).
In my case, I have been experiencing hang-ups when using a combination of EasyOCR and my custom libraries (using PyTorch >= 2.2.1 with Python 3.11, but maybe >= 2.2.0? with Python 3.12). However, it seems to be resolved by setting OMP_NUM_THREADS=1. I haven't conducted an exhaustive investigation, but anyway it has clearly improved the situation for me.
