I have a task to be solved. There are energy measurements over the square area 40x40. One measurement consists of values : x, y and the energy. The all area is almost whole covered with data (a few percent of cells are empty, we can consider them to have 0 energy). Many (thousands) measurements build one output state (a mask 40x40). We have 15000 of such outputs and their observations (measurements) - - that's a training set. The outcome should be a multi-value mask, also 40x40. It should be telling whether there is a matter and what type matter (different values of the mask). The mask can consist of a few figures, rectangularish, sometimes the figures are connected, they are all in the centre area of the output. Mask can have values 0-20, they are labels. Some sort of other specific information of the materials is given (density and radiation length). So instead of the labels we can use them (real numbers) to obtain resulting matter type. What DL model should I use?
I think you need a deep learning model that can perform both image segmentation and image classification. Image segmentation is the process of dividing an image into regions or pixels that share some common characteristics, such as color, intensity, or texture. Image classification is the process of assigning a label or class to each region or pixel, such as the type of matter.
One possible deep learning model that can do both tasks is a modified U-Net. A U-Net is a type of convolutional neural network that consists of an encoder and a decoder. The encoder extracts features from the input image, while the decoder reconstructs the output image with the desired labels.