rjcrystal

One discovery I've made with gemini is that ocr accuracy is much higher when document is perfectly aligned at 0 degree. When we provided images with handwritten text to gemini which were horizontal (90 or 180 degree) it had lots of issues reading dates, names etc. Then we used paddle ocr image orientation model to find orientation and rotate the image it solved most of our issues with ocr.

I work in healthcare domain, We've had great success converting printed lab reports (95%) to Json format using 1.5-Flash model. This post is really exciting for me. will definitely try out 2.0 models.

The struggle which almost every ocr usecase faces is with handwritten documents(doctor prescriptions with bad handwriting) With gemini 1.5 flash we've had ~75-80% percent accuracy (based on random sampling by pharmacists). we're planning to improve this further by fine-tuning gemini models with medical data.

What could be other alternative services/models for accurate handwriting ocr?

I love how fast everything is, it's the best searching tool, very low overhead needs little resources.

They're never in stock! They need to fix the logistical issues with supply.

I've been the software engineering game for the last 8+ and this is the best lesson I've learned in the past two years, been facing lot of health issues due to my workoholic nature and ignoring my physical health. Somehow my mental health has been good due to the support systems I have in my life and I'm super grateful for that.

Having a balanced life is my biggest goal right now as other carrer related things are mostly on auto pilot mode. I have a system set for that.

I just met an azure representative in a event and they said managed mysql is in private beta right now. It will take them around 3 months to get it out if it. Even I'm waiting for it

doesn't anyone think about performance anymore? I mean just add more vms or increase their config, no one talks about efficient resource utilization asfaik. And there might be times where you containerizing everything is an over overkill.

I saw some of the videos and i agree with the suggestions provided here but i really like the simple way of explaining and less mathematical more programmatic approach. It'd be awesome if you could make some neural networks with cuda or any other library like tensorflow or theano. Good luck.

I use a lenovo z500 3rd gen core i5 6 gigs of RAM and 250 gb samsung 850 evo ssd with Windows 7 and Linux mint. It's ssd runs like a charm. You gotta love the stability and simplicity of Windows 7 plus I experiment with stuffs on linux mint which is really awesome with mate as desktop environment and compiz window manager (you can do a lot of awesome window tweaks with it).