Is it possible to use FCIS: Fully Convolutional Instance-aware Semantic Segmentation without using CUDA?

I was reading a work on github (https://github.com/chrieke/InstanceSegmentation_Sentinel2) where it requires the use of FCIS and mxnet for deep learning, following the tutorial I saw that the installation of CUDA is required an extension of NVIDIA while I have available a video card AMD Radeon HD Series.

I wanted to know if I have trouble proceeding because I don’t have nvidia or can I use OpenCL as a better alternative?

I started with running the command sh ./init.sh but it returns:

skipping 'bbox.c' Cython extension (up-to-date)
running build_ext
Traceback (most recent call last):
  File "setup_linux.py", line 56, in <module>
  CUDA = locate_cuda()
File "setup_linux.py", line 51, in locate_cuda
  for k, v in cudaconfig.iteritems():
AttributeError: 'dict' object has no attribute 'iteritems'
Traceback (most recent call last):
File "setup_linux.py", line 56, in <module>
   CUDA = locate_cuda()
File "setup_linux.py", line 51, in locate_cuda
  for k, v in cudaconfig.iteritems():
AttributeError: 'dict' object has no attribute 'iteritems'

Secondly, running the command demo.py from FCIS says it doesn’t find the command even though the path is correct:

./FCS/fcis/demo.py
Traceback (most recent call last):
File "./fcis/demo.py", line 21, in
from utils.image import resize, transform
File "/home/..../..../lib/utils/image.py", line 6, in
from bbox.bbox_transform import clip_boxes
File "/home/..../..../lib/bbox/bbox_transform.py", line 3, in
from bbox import bbox_overlaps_cython
ImportError: cannot import name bbox_overlaps_cython

I changed it to from bbox import bbox_overlaps_cython to from bbox.bbox import bbox_overlaps_cython,and used python 2.7 as suggested but did not solve my problem.