example of general neural network: data, model, code, environment pip install tqdm
currently tested on two type of neural network models. One NN, and one PINN
the input should be a boundary picture with binary entry and fixed size
the output contain two part: first, whethe a pixel is a dot or not; second, whether two dots are connected.
This should be two NN. could refer to the linear regression
determine whether two dots are connected. But it is not needed in real case.
This should contain in loss function
n_dots_only_in_line
loss function = (n1 +n2)/n_dots_overlapping
for all pixels next to the dot, minimize
list of conditions:
- the dots must be on the boundary
- the dots are evenly distant with each other.
- how to adjust the dot density?
CNN (convolutional) network may be better for image processing.
it uses shared parameters and sparse connnections in graph.
Indeed, this img2dot issue only care about pixels in local region