Hello everyone,
I am working on an image recognition project and would like to ask for your help.
I am thinking of training a model using TensorFlow to read the instrument panel in the photo below.
Here are the steps I thought of:
1.) Verify if the image contains the equipment (Gea Dematron), as shown in the photo;
2.) Read the value on the panel (values range from 0.00 to 99.99).
I have already downloaded and installed the LabelImg software.
I would greatly appreciate it if you could give me some tips and suggestions for this project.
Something I’d suggest is that you use a pretrained model that is able to recognise digits for the numbers, and a model that is able to classify for the device.
Pretrained models have been trained in millions of images and will be more accurate, and you will need to do a lot less labelling of images.
The process of adapting a pretrained model to your task is called fine-tuning.
Now for reading out the digits it may be a bit more difficult, I’d take a look at some neural networks that are able to read characters in general, and try to feed only the red
area to the neural network.
Once you found some candidate models you can go back to us. But that’s what you need to find.