As you stated the input is an image. You can start with OCR as the first approach. For better results of text extraction and the postion of the text, I would use a local install of teserract engine.
Is doable technically.
Pre-process the image ( for example, convert it to a grayscale image and to make it high contrast)
Try all possible OCRs you can, UiPath, google, tesseract, tesseract local install, microsoft etc. (Here I am assuming the data is ok to be uploaded to cloud providers)
Most cloud providers of OCR will be quite accurate as there is only limited text in your sample image.
Use the position attributes if necessary. For example Bedroom can be found at different location. This can help you later when you want to check if the OCR did its job or not. Or to superimpose what the OCR found to a empty image without only text and no drawings.
Count number of Bedroom occurances
The output from Requirement 1 can also fetch you the dimensions of the rooms.
Calculate the area of the walls using the width length height of the walls.
Other than using OCR it’s not really an RPA case. In fact I think RPA is very limiting in getting a solution here.
Identifying the rooms and couting them should be doable with OCR, but the second point…
Assuming every piece of text can be neatly deriverd using some form OCR, and the dimensions derived as well, from that point on it becomes rather impossible without making assumptions.
You could gather the length and width of each room. Muliply them by 2 to get the total of all walls used, if each room had it’s own walls. This makes a rather big difference, since the lay-out of the entire structure becomes a factor. LEt alone the curved wall, diagonal walls, and wall extremities not belonging to the room.
Doors, passageways, windows. Are they considered ‘walls?’
the garage. It has no dimensions listed, nor is stated if it is included in the calculation. The answer also affects the wall-space calculation (inner vs outer, see point 1). Pars can be derived from the other rooms, but which is not easily solved by ‘RPA’.
And since so many of this is dependent on lay-out assumptions, it’s hard to standardize… making ‘automation’ quite obselete when it takes longer to build one bot then to do it manually.
Then again… I am old, so I might have missed an innovation here and there that can solve this…