HI,when the code release,i have waiting for a long time. about fcis HOT 17 OPEN

If the code is based on Caffe, could you publish the trainval.prototxt before the full code release? Maybe even the prototxt would answer many questions about module connectivity.

from fcis.

We have just released the code. We sincerely apologize for the delay. This is due to switching from our internal Caffe version to the public MXNet, which provides good support of fast multi-GPU training & inference.

Enjoy! It is worth noting that:

-FCIS provides a simple, fast and accurate framework for instance segmentation.

-Different from MNC, FCIS performs instance mask estimation and categorization jointly and simultaneously and estimates class-specific masks.

-We did not exploit the various techniques & tricks in the Mask RCNN system, like increasing RPN anchor numbers (from 12 to 15), enlarging the image (shorter side from 600 to 800 pixels), utilizing FPN features and aligned ROI pooling. These techniques & tricks should be orthogonal to our simple baseline.

from fcis.

Sorry for the delay. We will put all the effort in code release after ICCV deadline. Hopefully, it can be done in April.

from fcis.

I wonder who will first release the codes, Mask R-CNN or FCIS?

from fcis.

Hi - is there an update on the availability of the code. Even if just the caffe trained weights would be great!

from fcis.

At this moment it gives the following error.
xxx@deep:~/work/FCIS$ /home/xxx/anaconda2/bin/python ./fcis/demo.py
Traceback (most recent call last):
File "./fcis/demo.py", line 17, in
from utils.image import resize, transform
File "/home/xxx/work/FCIS/fcis/../lib/utils/image.py", line 6, in
from bbox.bbox_transform import clip_boxes
File "/home/xxx/work/FCIS/fcis/../lib/bbox/bbox_transform.py", line 11, in
from bbox import bbox_overlaps_cython
ImportError: cannot import name bbox_overlaps_cython

from fcis.

@alexeyda

Did you run sh ./init.sh to build the cython modules? (Following the installation instructions)

from fcis.

Is it possible to provide to a Docker image or a Dockerfile outlining steps involved?

I am working on it but might need couple of days.

from fcis.

Yes, I run it before. Hmm, just run it again and now it worked until the following step:
Traceback (most recent call last):
File "./fcis/demo.py", line 147, in
main()
File "./fcis/demo.py", line 43, in main
sym = sym_instance.get_symbol(config, is_train=False)
File "/home/jank/work/FCIS/fcis/symbols/resnet_v1_101_fcis.py", line 799, in get_symbol
psroipool_cls_seg = mx.contrib.sym.PSROIPooling(name='psroipool_cls_seg', data=fcis_cls_seg, rois=rois,
AttributeError: 'module' object has no attribute 'PSROIPooling'

from fcis.

Please make sure you installed mxnet with our operators correctly.
Check msracver/Deformable-ConvNets#8 for related discussion.

from fcis.

@alexeyda Hi I have met the same problem. Do you solve it? I'm sure I have followed the instruction.

from fcis.

@ZhengtianXu Make sure you have copied operator_cxx contents, not the directory itself, into mxnet/src/operators/contrib. i.e
cp -r ${YOUT_FCIS_ROOT}/fcis/operator_cxx/* ${YOUR_MXNET_ROOT}/src/operator/contrib/

from fcis.

@alexeyda hi, have you solved your problem?? I have the same error

from fcis.

@dbobrenko I followed what you have said, copy all files in operator_cxx to anaconda3/lib/python3.5/site-packages/mxnet/src/operator/contrib/, but it still doesn't work, by the way, how can I recompile mxnet?

from fcis.

@lc8631058 make sure you have compiled MXNET by yourself (not from pip), and copied these files before the MXNET compilation. For example:

1. Clone FCIS and MXNET repos as described in readme
2. Run bash init.sh in FCIS dir
3. Copy operators: cp ${YOUR_FCIS_ROOT}/fcis/operator_cxx/* ${YOUR_MXNET_ROOT}/src/operator/contrib/
4. Compile MXNET as described in their instructions

from fcis.

@dbobrenko hi, thanks a lot bro! now I run demo.py correctly in Jupyter notebook, but the show_masks function just output the original demo images without any changes, I wonder why?

from fcis.

@dbobrenko just found the answer in #21 ,thanks!

from fcis.