Daniel Lovera

Results 15 comments of Daniel Lovera

> The CRAFT detection model expects input images to be provided in RGB colorspace: https://github.com/clovaai/CRAFT-pytorch/blob/master/imgproc.py#L14 > > So it appears that EasyOCR is incorrect everywhere it is converting to BGR...

Also this section https://github.com/JaidedAI/EasyOCR/blob/c999505ef6b43be1c4ee36aa04ad979175178352/easyocr/utils.py#L566C1-L578C1 looks like it expects a RGB image as it's using PIL's resampling class.

Good find. Did you try changing the code so it passes an integer value of 4 to the openCV function? If so, did you see any performance differences? I'm going...

> I'm having a similar issue. Running on an Azure GPU worker - 8 cores, 56 GB RAM, and 16GB vRAM. It just shuts down. :( During training or inference?...

You can run just text detection. See the API documentation here: https://www.jaided.ai/easyocr/documentation/

I'd been meaning to add those stripe_customers definitions back to this repo, so I noticed you also added an id prop. Should that also be added to the creation of...

> if cp_section[-1]==[0,0], make last cy-c>cy+c. The final poly will be screwed I know this is years later, but any chance you can provide more context on what you meant...

I should clarify, this resolves GPU vRAM memory leaks. It's not resolving the CPU RAM memory leaks.

Corrected to only call empty_cache() if the device in use is cuda.

> The `del` stuff can't possibly work. It just removes the Python variable from the scope (the function) but doesn't actually remove anything from the GPU/CPU @jonashaag did you attempt...