Predict with pre-trained Faster RCNN models

I use mxnet-cu101==1.6.0b20190812 and gluoncv==0.4.0.post0 in Windows10.
I have a question about pre-trained Faster RCNN models of gluoncv.

from gluoncv import model_zoo, data, utils
im_fname =‘’ +
net = model_zoo.get_model(‘faster_rcnn_fpn_resnet101_v1d_coco’, pretrained=True)
x, img = data.transforms.presets.rcnn.load_test(im_fname)
class_IDs, scores, bounding_boxs = net(x)

An error occurred.

Traceback (most recent call last):
File “”, line 1, in
File “D:\python\mxnet0812\lib\site-packages\mxnet\gluon\”, line 548, in call
out = self.forward(*args)
File “D:\python\mxnet0812\lib\site-packages\mxnet\gluon\”, line 926, in forward
return self.hybrid_forward(ndarray, x, *args, **params)
File “D:\python\mxnet0812\lib\site-packages\gluoncv\model_zoo\faster_rcnn\”, line 344, in hybrid_forward
self._strides, roi_mode=self._roi_mode)
File “D:\python\mxnet0812\lib\site-packages\gluoncv\model_zoo\faster_rcnn\”, line 278, in _pyramid_roi_feats
TypeError: ROIAlign() takes from 0 to 2 positional arguments but 4 were given

When I used mxnet==1.4.1(cpu only), the error didn’t occur.

How do I change my code?
I would be grateful if anyone could tell me.

I updated to mxnet-cu101==1.6.0b20191004 and gluoncv==0.5.0.
The error disappeared.

@dai-ichiro, thanks for sharing your solution. Indeed specific version of gluon-cv are compatible with specific version of mxnet.