FloorNet
FloorNet copied to clipboard
art-programmer
Alternatives to Gurobi
Hello! It seems that Gurobi is being used which is not completely free. Any alternatives to evaluate the model without the use of Gurobi? It would have been better if this dependency is mentioned on the ReadMe file as well.
Sorry that I am not aware of free alternatives to Gurobi. I will keep you updated once we figure out how to get rid of Gurobi. I just updated the README file. Thanks for your suggestion!
@shubhamwagh I just replicate the IP optimization using a free solver. The interface is same as before. Please look at IP.py for more details.
That's really good. I will definitely take a look at it after holidays.
I was evaluating the performance of the trained model. I get an error
NUM_ICONS is not defined in IP.py
Is that the total number of icons in the icon folder?
Two things not included:
- from floorplan_utils.py import * -> is not included in IP.py script
- findMatches function is not included -> which I coped it from QP.py file
Hope this helps some one who will use it. @art-programmer it will be better if you can include it in your repository these things.
@art-programmer there are lot of bugs in IP.py like NUM_ROOMS has been used instead of NUM_FINAL_ROOMS and similarly for icons etc. This creates index pout of bounds errors. It will be better if you can rectify it.
Fixed the bugs, atleast I am able to evaluate the trained model with the provided dataset.
Fixed the bugs, atleast I am able to evaluate the trained model with the provided dataset.
Can you kindly share the bug-free IP.py file or guide me on how to solve the index pout of bounds errors?
This worked for me:
and in evaluate.py I have:
import numpy as np import cv2 #default #from QP import reconstructFloorplan, findMatches from QP import findMatches from IP import reconstructFloorplan from RecordReader import * from train import *
This worked for me:
and in evaluate.py I have:
import numpy as np import cv2 #default #from QP import reconstructFloorplan, findMatches from QP import findMatches from IP import reconstructFloorplan from RecordReader import * from train import *
Yes its working but u need install pulp 2.0 pip install pulp==2.0
