msracver / deep-feature-flow Goto Github PK

Dear Sir,

Can you kindly provide scripts for training and evaluation on cityscape dataset.'
I have tried but didn't understood how to end to end training on video sequence
I did trained end-to-end on key frame on cityscape dataset but end-to-end training on video sequence include key and current frame was issue. Hence requesting you kindly do the needful

Hi, @authors,

Thanks for releasing codes.

How can i begin a demo for video?

Thank you.

why learning rate in experiment/dff_rfcn/*.yaml is not as the learning rate as reported in paper?

Thank you for the code.I also look the FAQ you mentiond in the README.md
Q: It says AttributeError: 'module' object has no attribute 'MultiProposal'.
A: This is because either
you forget to copy the operators to your MXNet folder
or you copy to the wrong path
or you forget to re-compile and install
or you install the wrong MXNet
I followed every step mentioned in this README.md. But I also face the problem
--AttributeError: 'module' object has no attribute 'MultiProposal' when I run python ./rfcn/demo.

My PC: Ubuntu 16.04, CUDA 8, CUDNN 5.1,opencv3.2.0 ,python 2.7(the Ubuntu16.04 alreadly has).

I hope the answer,thanks!

Hi Orpine,

Thank you for solving the previous issues. When I train deep feature flow with r-fcn on imagenet dataset, I encounter this problem. could you please give a favor? waiting online.......

---------------------------------- error log below -----------------------------------------------------

ImageNetVID_DET_train_30classes gt roidb loaded from ./data/cache/ImageNetVID_DET_train_30classes_gt_roidb.pkl
append flipped images to roidb
num_images 57834
ImageNetVID_VID_train_15frames gt roidb loaded from ./data/cache/ImageNetVID_VID_train_15frames_gt_roidb.pkl
append flipped images to roidb
filtered 3316 roidb entries: 222946 -> 219630
providing maximum shape [('data', (1, 3, 600, 1000)), ('data_ref', (1, 3, 600, 1000)), ('eq_flag', (1,)), ('gt_boxes', (1, 100, 5))] [('label', (1, 21546)), ('bbox_target', (1, 36, 38, 63)), ('bbox_weight', (1, 36, 38, 63))]
{'bbox_target': (1L, 36L, 35L, 63L),
'bbox_weight': (1L, 36L, 35L, 63L),
'data': (1L, 3L, 556L, 1000L),
'data_ref': (1L, 3L, 556L, 1000L),
'eq_flag': (1L,),
'gt_boxes': (1L, 1L, 5L),
'im_info': (1L, 3L),
'label': (1L, 19845L)}
('lr', 0.00025, 'lr_epoch_diff', [1.333], 'lr_iters', [73191])
[14:46:31] src/operator/convolution.cu:87: This convolution is not supported by cudnn, MXNET convolution is applied.
[14:46:31] src/operator/convolution.cu:87: This convolution is not supported by cudnn, MXNET convolution is applied.
[14:46:31] src/operator/convolution.cu:87: This convolution is not supported by cudnn, MXNET convolution is applied.
[14:46:31] src/operator/convolution.cu:87: This convolution is not supported by cudnn, MXNET convolution is applied.
[14:46:31] /home/terry/mxnet/dmlc-core/include/dmlc/./logging.h:304: [14:46:31] /home/terry/mxnet/mshadow/mshadow/./tensor_gpu-inl.h:35: Check failed: e == cudaSuccess CUDA: invalid device ordinal

Stack trace returned 5 entries:
[bt] (0) /usr/local/lib/python2.7/dist-packages/mxnet-0.10.1-py2.7.egg/mxnet/libmxnet.so(_ZN4dmlc15LogMessageFatalD1Ev+0x3c) [0x7f065d6c6d7c]
[bt] (1) /usr/local/lib/python2.7/dist-packages/mxnet-0.10.1-py2.7.egg/mxnet/libmxnet.so(_ZNSt17_Function_handlerIFvvEZZN5mxnet6engine23ThreadedEnginePerDevice13PushToExecuteEPNS2_8OprBlockEbENKUlvE1_clEvEUlvE_E9_M_invokeERKSt9_Any_data+0x160) [0x7f065e48a650]
[bt] (2) /usr/lib/x86_64-linux-gnu/libstdc++.so.6(+0xb8c80) [0x7f0679c9fc80]
[bt] (3) /lib/x86_64-linux-gnu/libpthread.so.0(+0x76ba) [0x7f067c8826ba]
[bt] (4) /lib/x86_64-linux-gnu/libc.so.6(clone+0x6d) [0x7f067c5b83dd]

terminate called after throwing an instance of 'dmlc::Error'
what(): [14:46:31] /home/terry/mxnet/mshadow/mshadow/./tensor_gpu-inl.h:35: Check failed: e == cudaSuccess CUDA: invalid device ordinal

----------------------------------------error log above------------------------------------------------

Thanks!

[19:15:03] src/operator/convolution.cu:87: This convolution is not supported by cudnn, MXNET convolution is applied.
[19:15:03] src/operator/convolution.cu:87: This convolution is not supported by cudnn, MXNET convolution is applied.
[19:15:03] src/operator/convolution.cu:87: This convolution is not supported by cudnn, MXNET convolution is applied.
[19:15:03] src/operator/convolution.cu:87: This convolution is not supported by cudnn, MXNET convolution is applied.
[19:15:03] src/operator/convolution.cu:87: This convolution is not supported by cudnn, MXNET convolution is applied.
[19:15:03] src/operator/convolution.cu:87: This convolution is not supported by cudnn, MXNET convolution is applied.
[19:15:03] src/operator/convolution.cu:87: This convolution is not supported by cudnn, MXNET convolution is applied.
[19:15:03] src/operator/convolution.cu:87: This convolution is not supported by cudnn, MXNET convolution is applied.
[19:15:03] src/operator/convolution.cu:87: This convolution is not supported by cudnn, MXNET convolution is applied.
[19:15:03] src/operator/convolution.cu:87: This convolution is not supported by cudnn, MXNET convolution is applied.
[19:15:03] src/operator/convolution.cu:87: This convolution is not supported by cudnn, MXNET convolution is applied.
Traceback (most recent call last):
File "experiments/dff_rfcn/dff_rfcn_end2end_train_test.py", line 19, in
train_end2end.main()
File "experiments/dff_rfcn/../../dff_rfcn/train_end2end.py", line 171, in main
config.TRAIN.begin_epoch, config.TRAIN.end_epoch, config.TRAIN.lr, config.TRAIN.lr_step)
File "experiments/dff_rfcn/../../dff_rfcn/train_end2end.py", line 164, in train_net
arg_params=arg_params, aux_params=aux_params, begin_epoch=begin_epoch, num_epoch=end_epoch)
File "experiments/dff_rfcn/../../dff_rfcn/core/module.py", line 969, in fit
self.update()
File "experiments/dff_rfcn/../../dff_rfcn/core/module.py", line 1051, in update
self._curr_module.update()
File "experiments/dff_rfcn/../../dff_rfcn/core/module.py", line 572, in update
self._kvstore)
TypeError: _update_params_on_kvstore() takes exactly 4 arguments (3 given)

-------------------------------------------Error log above ------------------------------------------------

When I use DFF + R-FCN to train the imagenet dataset, DET+VID, it keep waing that "This convolution is not supported by cudnn, MXNET convolution is applied." and then error occur, is it because I am using Mxnet 0.10.0 release ?

Thank you.

Hi

When I try to clone the MXNet@(commit 62ecb60) branch using below command,

git clone --recursive -b 62ecb60 https://github.com/dmlc/mxnet

it shows fatal: Remote branch 62ecb60 not found, how can I get the MXNet@(commit 62ecb60) ? Have you tested on the latest mxnet 0.10.0 for Deep Feature Flow code ?

Thank you for your help.

Hi, all,

After I followed the steps in readme, I came across the following error when executing ./init.sh

(tf_1.0) root@milton-All-Series:/data/code/Deep-Feature-Flow# ./init.sh 
running build_ext
cythoning bbox.pyx to bbox.c
building 'bbox' extension
creating build
creating build/temp.linux-x86_64-3.5
gcc -pthread -Wno-unused-result -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -fPIC -I/root/anaconda3/envs/tf_1.0/lib/python3.5/site-packages/numpy/core/include -I/root/anaconda3/envs/tf_1.0/include/python3.5m -c bbox.c -o build/temp.linux-x86_64-3.5/bbox.o -Wno-cpp -Wno-unused-function
gcc -pthread -shared -L/root/anaconda3/envs/tf_1.0/lib -Wl,-rpath=/root/anaconda3/envs/tf_1.0/lib,--no-as-needed build/temp.linux-x86_64-3.5/bbox.o -L/root/anaconda3/envs/tf_1.0/lib -lpython3.5m -o /data/code/Deep-Feature-Flow/lib/bbox/bbox.cpython-35m-x86_64-linux-gnu.so
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'
(tf_1.0) root@milton-All-Series:/data/code/Deep-Feature-Flow#

Any suggestion to fix it?
I had installed: MXNet@(commit 62ecb60)
and Cython, opencv-python==3.2.0.6, easydict==1.6 as required.

Thanks~

When I run the demo, I met the problem:
Traceback (most recent call last):
File "./rfcn/demo.py", line 18, in
from utils.image import resize, transform
File "/home/Deep-Feature-Flow/rfcn/../lib/utils/image.py", line 6, in
from bbox.bbox_transform import clip_boxes
File "/home/Deep-Feature-Flow/rfcn/../lib/bbox/bbox_transform.py", line 2, in
from bbox import bbox_overlaps_cython
ImportError: cannot import name 'bbox_overlaps_cython'

Maybe it's a problem with cython, but I don't know how to solve it.
python and cython is with anaconda3

When I run dff_rfcn_end2end_train_test.py as instruction, it comes the error that in _update_params_on_kvstore() function lack a parameter.

Traceback (most recent call last):
File "experiments/dff_rfcn/dff_rfcn_end2end_train_test.py", line 19, in
train_end2end.main()
File "experiments/dff_rfcn/../../dff_rfcn/train_end2end.py", line 171, in main
config.TRAIN.begin_epoch, config.TRAIN.end_epoch, config.TRAIN.lr, config.TRAIN.lr_step)
File "experiments/dff_rfcn/../../dff_rfcn/train_end2end.py", line 164, in train_net
arg_params=arg_params, aux_params=aux_params, begin_epoch=begin_epoch, num_epoch=end_epoch)
File "experiments/dff_rfcn/../../dff_rfcn/core/module.py", line 974, in fit
self.update()
File "experiments/dff_rfcn/../../dff_rfcn/core/module.py", line 1056, in update
self._curr_module.update()
File "experiments/dff_rfcn/../../dff_rfcn/core/module.py", line 572, in update
self._kvstore
TypeError: _update_params_on_kvstore() takes exactly 4 arguments (3 given)

when I followed the Installation steps, I came across the opencv version problem here.

1. The official latest mxnet can be installed and run under python.

lucien@lucien-System-Product-Name:~$ python
Python 2.7.6 (default, Oct 26 2016, 20:30:19)
[GCC 4.8.4] on linux2
Type "help", "copyright", "credits" or "license" for more information.

import mxnet as mx
a = mx.nd.ones((2,3),mx.gpu())
b = a * 2 + 1
b.asnumpy()
array([[ 3., 3., 3.],
[ 3., 3., 3.]], dtype=float32)
exit()
lucien@lucien-System-Product-Name:~$ which python
/usr/bin/python
import mxnet
print mxnet.path
['/usr/local/lib/python2.7/dist-packages/mxnet']
print mxnet.version
0.9.5-1

but when run ./rfcn/demo.py, error occurred.

Traceback (most recent call last):
File "./rfcn/demo.py", line 142, in
main()
File "./rfcn/demo.py", line 49, in main
sym = sym_instance.get_test_symbol(config)
File "/home/lucien/Deep-Feature-Flow/rfcn/symbols/resnet_v1_101_rfcn.py", line 627, in get_test_symbol
rois = mx.contrib.sym.MultiProposal(
AttributeError: 'module' object has no attribute 'MultiProposal'

2. I went back to download the MXNet@(commit 62ecb60) and Copy operators in ./rfcn/operator_cxx to $(YOUR_MXNET_FOLDER)/src/operator/contrib and recompile MXNet.

/tmp/ccSXCcI0.o：在函数‘main’中：
im2rec.cc:(.text.startup+0x2e90)：对‘cv::imencode(std::__cxx11::basic_string<char, std::char_traits, std::allocator > const&, cv::_InputArray const&, std::vector<unsigned char, std::allocator >&, std::vector<int, std::allocator > const&)’未定义的引用
collect2: 错误：ld 返回 1
make: *** [bin/im2rec] 错误 1

so, I found it maybe opencv version problems.

lucien@lucien-System-Product-Name:/Deep-Feature-Flow$ pkg-config --libs opencv
/usr/local/lib/libopencv_calib3d.so /usr/local/lib/libopencv_contrib.so /usr/local/lib/libopencv_core.so /usr/local/lib/libopencv_features2d.so /usr/local/lib/libopencv_flann.so /usr/local/lib/libopencv_gpu.so /usr/local/lib/libopencv_highgui.so /usr/local/lib/libopencv_imgproc.so /usr/local/lib/libopencv_legacy.so /usr/local/lib/libopencv_ml.so /usr/local/lib/libopencv_nonfree.so /usr/local/lib/libopencv_objdetect.so /usr/local/lib/libopencv_ocl.so /usr/local/lib/libopencv_photo.so /usr/local/lib/libopencv_stitching.so /usr/local/lib/libopencv_superres.so /usr/local/lib/libopencv_ts.a /usr/local/lib/libopencv_video.so /usr/local/lib/libopencv_videostab.so /usr/lib/x86_64-linux-gnu/libXext.so /usr/lib/x86_64-linux-gnu/libX11.so /usr/lib/x86_64-linux-gnu/libICE.so /usr/lib/x86_64-linux-gnu/libSM.so /usr/lib/x86_64-linux-gnu/libGL.so /usr/lib/x86_64-linux-gnu/libGLU.so -L/usr/local/cuda/lib64 -L/usr/local/lib -lcufft -lcublas -lnpps -lnppi -lnppc -lcudart -ltbb -lrt -lpthread -lm -ldl -lopencv_calib3d -lopencv_contrib -lopencv_core -lopencv_features2d -lopencv_flann -lopencv_gpu -lopencv_highgui -lopencv_imgproc -lopencv_legacy -lopencv_ml -lopencv_nonfree -lopencv_objdetect -lopencv_ocl -lopencv_photo -lopencv_stitching -lopencv_superres -lopencv_ts -lopencv_video -lopencv_videostab
lucien@lucien-System-Product-Name:/Deep-Feature-Flow$ pkg-config --modversion opencv
2.4.10
lucien@lucien-System-Product-Name:~/Deep-Feature-Flow$ pip install opencv-python==3.2.0.6
Requirement already satisfied: opencv-python==3.2.0.6 in /usr/local/lib/python2.7/dist-packages
Requirement already satisfied: numpy>=1.11.3 in /usr/local/lib/python2.7/dist-packages (from opencv-python==3.2.0.6)

There is no module named 'libopencv_imgcodecs' under opencv 2.4.10. And opencv-python==3.2.0.6 has installed under python.

can I use the opencv-python 3.2.0.6 to recompile MXNet@(commit 62ecb60)?
Or I had to install opencv 3.2 version under '/usr/local/lib/', too? as the same as opencv 2.4.10?
I am confused. Waiting for your reply and suggestion.
p.s. such 'make clean & make' is useless answer, I had tried several times and did not work!!!

Ubuntu 14.04, Cuda 8 with cudnn 5.1, Python 2.7.6, GCC 4.8.4, single Titan X pascal card.

Hi all,

I try to apply other flownet to DFF. I need to confirm the image preprocess is the same.
However, I found the image is convert from BGR to RGB.

Second question.
I wonder what is the input of flownet.
Is that (image - pixel_mean)/255 ?
Some flownet repo normalize to [0, 1] and normalize again by imagenet mean, std
Normalize(mean=[0,0,0], std=[255,255,255])
Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
Why the preprocessing are quite different ?

Really need your help !
Thx !

HI,
when I set "ENABLE_OHEM:false" in the .yaml, the training goes gradient explosion. But I can reproduce the perfomance reported with "ENABLE_OHEM:true" .
Anyone knows the reason, or how to fix it?

with

Ubuntu 16.04LTS, geforce gtx 1060(6gb), cuda 8.0 cudnn 5.1

and Anaconda 2.

i installed mxnet with

$ pip install mxnet-cu80

and other requirements.

but there is any path like

(my mxnet source)src/operator/contrib

only has(used print mxnet.path in python)

/home/(name)/anaconda2/lib/python2.7/site-packages/mxnet

how could i recompile mxnet with new operator_cxx files????

plz help me :)

I'm trying to reproduce your experiments. However, I only have two gpus. In my experiments, the mAP on imagenet VID is only 69.9, which is lower than 73.1% in your paper. Is there anything to notice when change 4 gpus to 2 gpus.

I compiled mxnet with custom operator_cxx files and run demo

sudo python ./demo.py

and it prints out

/usr/lib/python2.7/dist-packages/matplotlib/font_manager.py:273: UserWarning: Matplotlib is building the font cache using fc-list. This may take a moment.
warnings.warn('Matplotlib is building the font cache using fc-list. This may take a moment.')
{'CLASS_AGNOSTIC': True,
'MXNET_VERSION': 'mxnet',
'SCALES': [(600, 1000)],
'TEST': {'BATCH_IMAGES': 1,
'CXX_PROPOSAL': True,
'HAS_RPN': True,
'NMS': 0.3,
'RPN_MIN_SIZE': 0,
'RPN_NMS_THRESH': 0.7,
'RPN_POST_NMS_TOP_N': 300,
'RPN_PRE_NMS_TOP_N': 6000,
'max_per_image': 300,
'test_epoch': 2},
'TRAIN': {'ASPECT_GROUPING': True,
'BATCH_IMAGES': 1,
'BATCH_ROIS': -1,
'BATCH_ROIS_OHEM': 128,
'BBOX_MEANS': [0.0, 0.0, 0.0, 0.0],
'BBOX_NORMALIZATION_PRECOMPUTED': True,
'BBOX_REGRESSION_THRESH': 0.5,
'BBOX_STDS': [0.1, 0.1, 0.2, 0.2],
'BBOX_WEIGHTS': array([ 1., 1., 1., 1.]),
'BG_THRESH_HI': 0.5,
'BG_THRESH_LO': 0.0,
'CXX_PROPOSAL': True,
'ENABLE_OHEM': True,
'END2END': True,
'FG_FRACTION': 0.25,
'FG_THRESH': 0.5,
'FLIP': True,
'RESUME': False,
'RPN_BATCH_SIZE': 256,
'RPN_BBOX_WEIGHTS': [1.0, 1.0, 1.0, 1.0],
'RPN_CLOBBER_POSITIVES': False,
'RPN_FG_FRACTION': 0.5,
'RPN_MIN_SIZE': 0,
'RPN_NEGATIVE_OVERLAP': 0.3,
'RPN_NMS_THRESH': 0.7,
'RPN_POSITIVE_OVERLAP': 0.7,
'RPN_POSITIVE_WEIGHT': -1.0,
'RPN_POST_NMS_TOP_N': 300,
'RPN_PRE_NMS_TOP_N': 6000,
'SHUFFLE': True,
'begin_epoch': 0,
'end_epoch': 2,
'lr': 0.00025,
'lr_factor': 0.1,
'lr_step': '1.333',
'model_prefix': 'rfcn_vid',
'momentum': 0.9,
'warmup': False,
'warmup_lr': 0,
'warmup_step': 0,
'wd': 0.0005},
'dataset': {'NUM_CLASSES': 31,
'dataset': 'ImageNetVID',
'dataset_path': './data/ILSVRC2015',
'image_set': 'DET_train_30classes+VID_train_15frames',
'proposal': 'rpn',
'root_path': './data',
'test_image_set': 'VID_val_frames'},
'default': {'frequent': 100, 'kvstore': 'device'},
'gpus': '0',
'network': {'ANCHOR_MEANS': [0.0, 0.0, 0.0, 0.0],
'ANCHOR_RATIOS': [0.5, 1, 2],
'ANCHOR_SCALES': [8, 16, 32],
'ANCHOR_STDS': [0.1, 0.1, 0.4, 0.4],
'FIXED_PARAMS': ['conv1',
'bn_conv1',
'res2',
'bn2',
'gamma',
'beta'],
'IMAGE_STRIDE': 0,
'NORMALIZE_RPN': True,
'NUM_ANCHORS': 9,
'PIXEL_MEANS': array([ 103.06, 115.9 , 123.15]),
'RCNN_FEAT_STRIDE': 16,
'RPN_FEAT_STRIDE': 16,
'pretrained': '',
'pretrained_epoch': 0},
'output_path': './output/rfcn/imagenet_vid',
'symbol': ''}
[17:42:26] /home/han/mxnet/dmlc-core/include/dmlc/./logging.h:300: [17:42:25] src/c_api/c_api_ndarray.cc:390: Operator _zeros cannot be run; requires at least one of FCompute, NDArrayFunction, FCreateOperator be registered

Stack trace returned 10 entries:
[bt] (0) /usr/local/lib/python2.7/dist-packages/mxnet-0.9.5-py2.7.egg/mxnet/libmxnet.so(_ZN4dmlc15LogMessageFatalD1Ev+0x3c) [0x7fd683023f9c]
[bt] (1) /usr/local/lib/python2.7/dist-packages/mxnet-0.9.5-py2.7.egg/mxnet/libmxnet.so(MXImperativeInvoke+0x8d1) [0x7fd683bb4951]
[bt] (2) /usr/lib/x86_64-linux-gnu/libffi.so.6(ffi_call_unix64+0x4c) [0x7fd6865c3e40]
[bt] (3) /usr/lib/x86_64-linux-gnu/libffi.so.6(ffi_call+0x2eb) [0x7fd6865c38ab]
[bt] (4) /usr/lib/python2.7/lib-dynload/_ctypes.x86_64-linux-gnu.so(_ctypes_callproc+0x48f) [0x7fd6867d33df]
[bt] (5) /usr/lib/python2.7/lib-dynload/_ctypes.x86_64-linux-gnu.so(+0x11d82) [0x7fd6867d7d82]
[bt] (6) python(PyObject_Call+0x43) [0x4b0cb3]
[bt] (7) python(PyEval_EvalFrameEx+0x5faf) [0x4c9faf]
[bt] (8) python(PyEval_EvalCodeEx+0x255) [0x4c2765]
[bt] (9) python(PyEval_EvalFrameEx+0x68d1) [0x4ca8d1]

Traceback (most recent call last):
File "./demo.py", line 142, in
main()
File "./demo.py", line 93, in main
arg_params=arg_params, aux_params=aux_params)
File "/home/han/Deep-Feature-Flow/rfcn/core/tester.py", line 29, in init
self._mod.bind(provide_data, provide_label, for_training=False)
File "/home/han/Deep-Feature-Flow/rfcn/core/module.py", line 839, in bind
for_training, inputs_need_grad, force_rebind=False, shared_module=None)
File "/home/han/Deep-Feature-Flow/rfcn/core/module.py", line 396, in bind
state_names=self._state_names)
File "/home/han/Deep-Feature-Flow/rfcn/core/DataParallelExecutorGroup.py", line 186, in init
self.bind_exec(data_shapes, label_shapes, shared_group)
File "/home/han/Deep-Feature-Flow/rfcn/core/DataParallelExecutorGroup.py", line 272, in bind_exec
shared_group))
File "/home/han/Deep-Feature-Flow/rfcn/core/DataParallelExecutorGroup.py", line 545, in _bind_ith_exec
context, self.logger)
File "/home/han/Deep-Feature-Flow/rfcn/core/DataParallelExecutorGroup.py", line 523, in _get_or_reshape
arg_arr = nd.zeros(arg_shape, context, dtype=arg_type)
File "/usr/local/lib/python2.7/dist-packages/mxnet-0.9.5-py2.7.egg/mxnet/ndarray.py", line 946, in zeros
return _internal._zeros(shape=shape, ctx=ctx, dtype=dtype)
File "/usr/local/lib/python2.7/dist-packages/mxnet-0.9.5-py2.7.egg/mxnet/_ctypes/ndarray.py", line 164, in generic_ndarray_function
c_array(ctypes.c_char_p, [c_str(val) for val in vals])))
File "/usr/local/lib/python2.7/dist-packages/mxnet-0.9.5-py2.7.egg/mxnet/base.py", line 78, in check_call
raise MXNetError(py_str(_LIB.MXGetLastError()))
mxnet.base.MXNetError: [17:42:25] src/c_api/c_api_ndarray.cc:390: Operator _zeros cannot be run; requires at least one of FCompute, NDArrayFunction, FCreateOperator be registered

Stack trace returned 10 entries:
[bt] (0) /usr/local/lib/python2.7/dist-packages/mxnet-0.9.5-py2.7.egg/mxnet/libmxnet.so(_ZN4dmlc15LogMessageFatalD1Ev+0x3c) [0x7fd683023f9c]
[bt] (1) /usr/local/lib/python2.7/dist-packages/mxnet-0.9.5-py2.7.egg/mxnet/libmxnet.so(MXImperativeInvoke+0x8d1) [0x7fd683bb4951]
[bt] (2) /usr/lib/x86_64-linux-gnu/libffi.so.6(ffi_call_unix64+0x4c) [0x7fd6865c3e40]
[bt] (3) /usr/lib/x86_64-linux-gnu/libffi.so.6(ffi_call+0x2eb) [0x7fd6865c38ab]
[bt] (4) /usr/lib/python2.7/lib-dynload/_ctypes.x86_64-linux-gnu.so(_ctypes_callproc+0x48f) [0x7fd6867d33df]
[bt] (5) /usr/lib/python2.7/lib-dynload/_ctypes.x86_64-linux-gnu.so(+0x11d82) [0x7fd6867d7d82]
[bt] (6) python(PyObject_Call+0x43) [0x4b0cb3]
[bt] (7) python(PyEval_EvalFrameEx+0x5faf) [0x4c9faf]
[bt] (8) python(PyEval_EvalCodeEx+0x255) [0x4c2765]
[bt] (9) python(PyEval_EvalFrameEx+0x68d1) [0x4ca8d1]

I can't find even where error is occurred!
is that error occurred from mxnet???

plz help me friends :(

i ran this code on

ubuntu 16.04lts, with cuda8.0 cudnn5.1

Hi, @authors,

Thanks for releasing this package.
When would you release the demo on segmentation for cityscapes dataset? Currently I only see the demo on (ImageNet DET + VID train).

Thanks!

Hi,

'output_path': './output/rfcn/imagenet_vid',
'symbol': ''}
Traceback (most recent call last):
File "./rfcn/demo.py", line 142, in
main()
File "./rfcn/demo.py", line 73, in main
im = cv2.imread(im_name, cv2.IMREAD_COLOR | cv2.IMREAD_IGNORE_ORIENTATION)
AttributeError: 'module' object has no attribute 'IMREAD_IGNORE_ORIENTATION'

I do have copies the four files under operator_cxx to (MXNET_FOLDER)/src/operator/contrib

rfcn/operator_cxx
|-------multi_proposal.cc multi_proposal-inl.h multi_proposal.cu psroi_pooling.cc psroi_pooling-inl.h psroi_pooling.cc psroi_pooling.cu

mxnet/src/operator/contrib
|-------multi_proposal.cc multi_proposal-inl.h multi_proposal.cu psroi_pooling.cc psroi_pooling-inl.h psroi_pooling.cc psroi_pooling.cu

and recompile mxnet: make -j $(nproc) USE_OPENCV=1 USE_BLAS=openblas USE_CUDA=1 USE_CUDA_PATH=/usr/local/cuda USE_CUDNN=1

and install in (MXNET_FOLDER)/python: sudo python setup.py install.

The error shows when I run the demo. I have tried the same on another server, the same error shows. How to solve this problem? Thank you.

I just run the comman:
python ./rfcn/demo.py

what i found is that it can't detect person,it's strange,so this module didn't provide the model to detect person?If I want to detect person,how should I do?

I met problem when I trying the demo python ./rfcn/demo.py

platform: ubuntu 16.04

`[10:21:13] /home/zhouyang/github/mxnet/dmlc-core/include/dmlc/./logging.h:300: [10:21:13] src/c_api/c_api_ndarray.cc:390: Operator _zeros cannot be run; requires at least one of FCompute, NDArrayFunction, FCreateOperator be registered

Stack trace returned 10 entries:
[bt] (0) /home/zhouyang/anaconda2/lib/python2.7/site-packages/mxnet-0.9.5-py2.7.egg/mxnet/libmxnet.so(_ZN4dmlc15LogMessageFatalD1Ev+0x3c) [0x7fe5ee40214c]
[bt] (1) /home/zhouyang/anaconda2/lib/python2.7/site-packages/mxnet-0.9.5-py2.7.egg/mxnet/libmxnet.so(MXImperativeInvoke+0x8d1) [0x7fe5eefabc61]
[bt] (2) /home/zhouyang/anaconda2/lib/python2.7/lib-dynload/_ctypes.so(ffi_call_unix64+0x4c) [0x7fe5f7a3757c]
[bt] (3) /home/zhouyang/anaconda2/lib/python2.7/lib-dynload/_ctypes.so(ffi_call+0x1f5) [0x7fe5f7a36cd5]
[bt] (4) /home/zhouyang/anaconda2/lib/python2.7/lib-dynload/_ctypes.so(_ctypes_callproc+0x3e6) [0x7fe5f7a2e376]
[bt] (5) /home/zhouyang/anaconda2/lib/python2.7/lib-dynload/_ctypes.so(+0x9db3) [0x7fe5f7a25db3]
[bt] (6) /home/zhouyang/anaconda2/bin/../lib/libpython2.7.so.1.0(PyObject_Call+0x53) [0x7fe5fc58ae93]
[bt] (7) /home/zhouyang/anaconda2/bin/../lib/libpython2.7.so.1.0(PyEval_EvalFrameEx+0x715d) [0x7fe5fc63d80d]
[bt] (8) /home/zhouyang/anaconda2/bin/../lib/libpython2.7.so.1.0(PyEval_EvalCodeEx+0x89e) [0x7fe5fc63fc3e]
[bt] (9) /home/zhouyang/anaconda2/bin/../lib/libpython2.7.so.1.0(PyEval_EvalFrameEx+0x8b47) [0x7fe5fc63f1f7]

Traceback (most recent call last):
File "./rfcn/demo.py", line 142, in
main()
File "./rfcn/demo.py", line 93, in main
arg_params=arg_params, aux_params=aux_params)
File "/home/zhouyang/github/Deep-Feature-Flow/rfcn/core/tester.py", line 29, in init
self._mod.bind(provide_data, provide_label, for_training=False)
File "/home/zhouyang/github/Deep-Feature-Flow/rfcn/core/module.py", line 839, in bind
for_training, inputs_need_grad, force_rebind=False, shared_module=None)
File "/home/zhouyang/github/Deep-Feature-Flow/rfcn/core/module.py", line 396, in bind
state_names=self._state_names)
File "/home/zhouyang/github/Deep-Feature-Flow/rfcn/core/DataParallelExecutorGroup.py", line 186, in init
self.bind_exec(data_shapes, label_shapes, shared_group)
File "/home/zhouyang/github/Deep-Feature-Flow/rfcn/core/DataParallelExecutorGroup.py", line 272, in bind_exec
shared_group))
File "/home/zhouyang/github/Deep-Feature-Flow/rfcn/core/DataParallelExecutorGroup.py", line 545, in _bind_ith_exec
context, self.logger)
File "/home/zhouyang/github/Deep-Feature-Flow/rfcn/core/DataParallelExecutorGroup.py", line 523, in _get_or_reshape
arg_arr = nd.zeros(arg_shape, context, dtype=arg_type)
File "/home/zhouyang/anaconda2/lib/python2.7/site-packages/mxnet-0.9.5-py2.7.egg/mxnet/ndarray.py", line 946, in zeros
return _internal._zeros(shape=shape, ctx=ctx, dtype=dtype)
File "/home/zhouyang/anaconda2/lib/python2.7/site-packages/mxnet-0.9.5-py2.7.egg/mxnet/_ctypes/ndarray.py", line 164, in generic_ndarray_function
c_array(ctypes.c_char_p, [c_str(val) for val in vals])))
File "/home/zhouyang/anaconda2/lib/python2.7/site-packages/mxnet-0.9.5-py2.7.egg/mxnet/base.py", line 78, in check_call
raise MXNetError(py_str(_LIB.MXGetLastError()))
mxnet.base.MXNetError: [10:21:13] src/c_api/c_api_ndarray.cc:390: Operator _zeros cannot be run; requires at least one of FCompute, NDArrayFunction, FCreateOperator be registered

Stack trace returned 10 entries:
[bt] (0) /home/zhouyang/anaconda2/lib/python2.7/site-packages/mxnet-0.9.5-py2.7.egg/mxnet/libmxnet.so(_ZN4dmlc15LogMessageFatalD1Ev+0x3c) [0x7fe5ee40214c]
[bt] (1) /home/zhouyang/anaconda2/lib/python2.7/site-packages/mxnet-0.9.5-py2.7.egg/mxnet/libmxnet.so(MXImperativeInvoke+0x8d1) [0x7fe5eefabc61]
[bt] (2) /home/zhouyang/anaconda2/lib/python2.7/lib-dynload/_ctypes.so(ffi_call_unix64+0x4c) [0x7fe5f7a3757c]
[bt] (3) /home/zhouyang/anaconda2/lib/python2.7/lib-dynload/_ctypes.so(ffi_call+0x1f5) [0x7fe5f7a36cd5]
[bt] (4) /home/zhouyang/anaconda2/lib/python2.7/lib-dynload/_ctypes.so(_ctypes_callproc+0x3e6) [0x7fe5f7a2e376]
[bt] (5) /home/zhouyang/anaconda2/lib/python2.7/lib-dynload/_ctypes.so(+0x9db3) [0x7fe5f7a25db3]
[bt] (6) /home/zhouyang/anaconda2/bin/../lib/libpython2.7.so.1.0(PyObject_Call+0x53) [0x7fe5fc58ae93]
[bt] (7) /home/zhouyang/anaconda2/bin/../lib/libpython2.7.so.1.0(PyEval_EvalFrameEx+0x715d) [0x7fe5fc63d80d]
[bt] (8) /home/zhouyang/anaconda2/bin/../lib/libpython2.7.so.1.0(PyEval_EvalCodeEx+0x89e) [0x7fe5fc63fc3e]
[bt] (9) /home/zhouyang/anaconda2/bin/../lib/libpython2.7.so.1.0(PyEval_EvalFrameEx+0x8b47) [0x7fe5fc63f1f7]`

Hi, thank you for releasing this code. However I cannot run it as expected.

Traceback (most recent call last):
  File "/home/cory/Deep-Feature-Flow/rfcn/demo.py", line 142, in <module>
    main()
  File "/home/cory/Deep-Feature-Flow/rfcn/demo.py", line 49, in main
    sym = sym_instance.get_test_symbol(config)
  File "/home/cory/Deep-Feature-Flow/rfcn/symbols/resnet_v1_101_rfcn.py", line 628, in get_test_symbol
    rois = mx.contrib.sym.MultiProposal(
AttributeError: 'module' object has no attribute 'MultiProposal'

I followed every step mentioned in this repo. Recompiling MXNet with operator_cxx worked very well. But in python code it seems not found these ops.

However, I can build and run Deformable-ConvNets without any error. The building processes of these 2 repos are almost the same. Can you please give me some hint ?

My PC: Ubuntu 16.04, Anaconda 2, CUDA 8, CUDNN 5.1

[09:38:23] /home/pdl/workspace2/ylj/MXNet/mxnet/dmlc-core/include/dmlc/logging.h:308: [09:38:23] src/operator/batch_norm-inl.h:238: Check failed: channelAxis < dshape.ndim() (1 vs. 0) Channel axis out of range: 1
Stack trace returned 10 entries:
[bt] (0) /home/pdl/workspace2/ylj/MXNet/mxnet/python/mxnet/../../lib/libmxnet.so(_ZN4dmlc15LogMessageFatalD1Ev+0x3c) [0x7fd8fec99aac]
[bt] (1) /home/pdl/workspace2/ylj/MXNet/mxnet/python/mxnet/../../lib/libmxnet.so(ZNK5mxnet2op13BatchNormProp10InferShapeEPSt6vectorIN4nnvm6TShapeESaIS4_EES7_S7+0x979) [0x7fd8ffbab989]
[bt] (2) /home/pdl/workspace2/ylj/MXNet/mxnet/python/mxnet/../../lib/libmxnet.so(+0x16979e7) [0x7fd8ffb719e7]
[bt] (3) /home/pdl/workspace2/ylj/MXNet/mxnet/python/mxnet/../../lib/libmxnet.so(+0x152be47) [0x7fd8ffa05e47]
[bt] (4) /home/pdl/workspace2/ylj/MXNet/mxnet/python/mxnet/../../lib/libmxnet.so(_ZN5mxnet4exec10InferShapeEN4nnvm5GraphESt6vectorINS1_6TShapeESaIS4_EERKSs+0x83b) [0x7fd8ffa07afb]
[bt] (5) /home/pdl/workspace2/ylj/MXNet/mxnet/python/mxnet/../../lib/libmxnet.so(MXSymbolInferShape+0x17ed) [0x7fd8ff99f62d]
[bt] (6) /home/pdl/anaconda2/lib/python2.7/lib-dynload/_ctypes.so(ffi_call_unix64+0x4c) [0x7fd90e97857c]
[bt] (7) /home/pdl/anaconda2/lib/python2.7/lib-dynload/_ctypes.so(ffi_call+0x1f5) [0x7fd90e977cd5]
[bt] (8) /home/pdl/anaconda2/lib/python2.7/lib-dynload/_ctypes.so(_ctypes_callproc+0x3e6) [0x7fd90e96f376]
[bt] (9) /home/pdl/anaconda2/lib/python2.7/lib-dynload/_ctypes.so(+0x9db3) [0x7fd90e966db3]

infer_shape error. Arguments:
data: (1L, 3L, 562L, 1000L)
label: (1L, 20412L)
bbox_target: (1L, 36L, 36L, 63L)
bbox_weight: (1L, 36L, 36L, 63L)
Traceback (most recent call last):
File "dff_rfcn_end2end_train_test.py", line 19, in
train_end2end.main()
File "../../dff_rfcn/train_end2end.py", line 182, in main
config['TRAIN']['begin_epoch'], config['TRAIN']['end_epoch'], config['TRAIN']['lr'], config['TRAIN']['lr_step'])
File "../../dff_rfcn/train_end2end.py", line 101, in train_net
sym_instance.infer_shape(data_shape_dict)
File "../../dff_rfcn/../lib/utils/symbol.py", line 38, in infer_shape
arg_shape, out_shape, aux_shape = self.sym.infer_shape(**data_shape_dict)
File "/home/pdl/workspace2/ylj/MXNet/mxnet/python/mxnet/symbol/symbol.py", line 958, in infer_shape
res = self._infer_shape_impl(False, *args, **kwargs)
File "/home/pdl/workspace2/ylj/MXNet/mxnet/python/mxnet/symbol/symbol.py", line 1087, in _infer_shape_impl
ctypes.byref(complete)))
File "/home/pdl/workspace2/ylj/MXNet/mxnet/python/mxnet/base.py", line 143, in check_call
raise MXNetError(py_str(_LIB.MXGetLastError()))
mxnet.base.MXNetError: Error in operator bn_conv1: [09:38:23] src/operator/batch_norm-inl.h:238: Check failed: channelAxis < dshape.ndim() (1 vs. 0) Channel axis out of range: 1

Stack trace returned 10 entries:
[bt] (0) /home/pdl/workspace2/ylj/MXNet/mxnet/python/mxnet/../../lib/libmxnet.so(_ZN4dmlc15LogMessageFatalD1Ev+0x3c) [0x7fd8fec99aac]
[bt] (1) /home/pdl/workspace2/ylj/MXNet/mxnet/python/mxnet/../../lib/libmxnet.so(ZNK5mxnet2op13BatchNormProp10InferShapeEPSt6vectorIN4nnvm6TShapeESaIS4_EES7_S7+0x979) [0x7fd8ffbab989]
[bt] (2) /home/pdl/workspace2/ylj/MXNet/mxnet/python/mxnet/../../lib/libmxnet.so(+0x16979e7) [0x7fd8ffb719e7]
[bt] (3) /home/pdl/workspace2/ylj/MXNet/mxnet/python/mxnet/../../lib/libmxnet.so(+0x152be47) [0x7fd8ffa05e47]
[bt] (4) /home/pdl/workspace2/ylj/MXNet/mxnet/python/mxnet/../../lib/libmxnet.so(_ZN5mxnet4exec10InferShapeEN4nnvm5GraphESt6vectorINS1_6TShapeESaIS4_EERKSs+0x83b) [0x7fd8ffa07afb]
[bt] (5) /home/pdl/workspace2/ylj/MXNet/mxnet/python/mxnet/../../lib/libmxnet.so(MXSymbolInferShape+0x17ed) [0x7fd8ff99f62d]
[bt] (6) /home/pdl/anaconda2/lib/python2.7/lib-dynload/_ctypes.so(ffi_call_unix64+0x4c) [0x7fd90e97857c]
[bt] (7) /home/pdl/anaconda2/lib/python2.7/lib-dynload/_ctypes.so(ffi_call+0x1f5) [0x7fd90e977cd5]
[bt] (8) /home/pdl/anaconda2/lib/python2.7/lib-dynload/_ctypes.so(_ctypes_callproc+0x3e6) [0x7fd90e96f376]
[bt] (9) /home/pdl/anaconda2/lib/python2.7/lib-dynload/_ctypes.so(+0x9db3) [0x7fd90e966db3]

It seems like something wrong with the image shape， but I don't know how to solve it. Help， please！

hi, i want to make a caffe version about warping layer, can you give me some suggestion?

hi,all
i want to try this algrithm on virtual machine and my system is ubuntu 16.04.
I had installed: MXNet@(commit 62ecb60) and Cython, opencv-python==3.2.0.6, easydict==1.6 as required. My python version is 2.7.12. But i came across this problem when i run init.sh

Traceback (most recent call last):
File "setup_linux.py", line 56, in
CUDA = locate_cuda()
File "setup_linux.py", line 44, in locate_cuda
raise EnvironmentError('The nvcc binary could not be '
EnvironmentError: The nvcc binary could not be located in your $PATH. Either add it to your path, or set $CUDAHOME

how can i fix this problem? is it possible to run the demo on virtual machine?

I cannot find any code example for video segmentation. Anyone can help me?

Thanks for sharing the code. Got a little question.
Nowhere could I find the setting for the key frame duration length in the test code or cfg file.
Is it done in the process of extracting validation RoIdb?

Beacause of some reason, OneDrive can not use in China normally.
Please provide another way ( like BaiduYun ) to download the pretrained model !
Thank you !

Hi, @daijifeng001
Thanks for releasing this package.
I got into trouble when I tried to run the demo 'python ./rfcn/demo.py',and it is like that **AttributeError: 'module' object has no attribute 'MultiProposal' **.

I have copied operators in ./rfcn/operator_cxx to $(YOUR_MXNET_FOLDER)/src/operator/contrib and recompiled MXNet successfully. By the way, I copied just like this, cp ./rfcn/operator_cxx/* $(YOUR_MXNET_FOLDER)/src/operator/contrib/*, and I also tried to copy the folder, cp -rf ./rfcn/operator_cxx $(YOUR_MXNET_FOLDER)/src/operator/contrib .

So, does anything go wrong? What should I do?
Thanks.

Hi,

Thanks for the share of code.

I have written Dockerfile for this repository, supporting CUDA. The address is https://github.com/ZHAOZHIHAO/RunningProgramms/tree/master/running_deep_feature_flow_for_video_recognition
, readme inside.

I hope it can help someone.

Best,
Zhihao

Hi, thx for sharing your great job.
I want to evaluate your inference , can you show me how to use the evaluation code?

I read the R-FCN paper.

After the Resnet101 they append a randomly initialized 1024-d 1x1 convolution layer to reducing dimension.
However, your implementation append the a 1024-d "3x3" dilation="6" convolution layer.
The paper doesn't discuss this difference.
I wondering if I use 1024-d 1x1 convolution layer like R-FCN, whether the result will be different or not?

Thx for your help !

Such as Baidu Yun Driver

https://github.com/msracver/Deep-Feature-Flow/blob/master/dff_rfcn/symbols/resnet_v1_101_flownet_rfcn.py#L539

I wondering why multiplying 2.5 to the output of flownet?
I didn't see any description in paper.
Thanks for your help!

The OneDrive link with pretrained models does not contain the Flownet model (only the Resnet model). A link to the pretrained Flownet model would be highly appreciated!

while i follow mxnet installation

3.1 Clone MXNet and checkout to MXNet@(commit 62ecb60) by

git clone --recursive https://github.com/dmlc/mxnet.git
git checkout 62ecb60
git submodule update

after

git submodule update

it prints out

fatal: reference is not a tree: 89de7ab20167909bc2c4f8acd397671c47cf3c0d
Unable to checkout '89de7ab20167909bc2c4f8acd397671c47cf3c0d' in submodule path 'cub'

if i ignore that fatal error, then another error occurs during make operation

make -j8

it prints out

/usr/bin/ld: cannot find -lcblas
collect2: error: ld returned 1 exit status
Makefile:242: recipe for target 'lib/libmxnet.so' failed
make: *** [lib/libmxnet.so] Error 1
make: *** Waiting for unfinished jobs....
/usr/bin/ld: cannot find -lcblas
collect2: error: ld returned 1 exit status
Makefile:264: recipe for target 'bin/im2rec' failed
make: *** [bin/im2rec] Error 1

please help me guyz

Hi, @daijifeng001 @orpine
Thanks for releasing this package.
I'd like to ask how can you tune this model on cityscapes dataset with sparse label for segmentation? As I know, the dataset only give a label every 30 frames. Thanks for your help!

Hi,

When I train and test Deep Feature Flow with R-FCN, using the following command:
python experiments/dff_rfcn/dff_rfcn_end2end_train_test.py --cfg experiments/dff_rfcn/cfgs/resnet_v1_101_flownet_imagenet_vid_rfcn_end2end_ohem.yaml

AssertionError: Path does not exist: ./data/ILSVRC2015/ImageSets/DET_train_30classes.txt

Is it only 30 classes information for VID in DET_train_30classes.txt, what is the calss data format inside txt file? Or where can I download the DET_train_30calsses.txt ??

Thank you !!!

I am running into the following error when trying to run the first demo. I have followed all the steps in the README until that point (MXNet built successfully, etc.)

root@ip-172-31-15-137:/ebs/Deep-Feature-Flow# python2 ./rfcn/demo.py
libdc1394 error: Failed to initialize libdc1394

(demo.py:22616): Gtk-ERROR **: GTK+ 2.x symbols detected. Using GTK+ 2.x and GTK+ 3 in the same process is not supported
Trace/breakpoint trap

hi,
I'm running the demo and here is the error I encountered:
`[10:46:34] /home/huan/Deep-Feature-Flow/mxnet/dmlc-core/include/dmlc/./logging.h:300: [10:46:34] src/c_api/c_api_ndarray.cc:390: Operator _zeros cannot be run; requires at least one of FCompute, NDArrayFunction, FCreateOperator be registered

Stack trace returned 10 entries:
[bt] (0) /usr/local/lib/python2.7/dist-packages/mxnet-0.9.5-py2.7.egg/mxnet/libmxnet.so(_ZN4dmlc15LogMessageFatalD1Ev+0x3c) [0x7f0d5982bf9c]
[bt] (1) /usr/local/lib/python2.7/dist-packages/mxnet-0.9.5-py2.7.egg/mxnet/libmxnet.so(MXImperativeInvoke+0x8d1) [0x7f0d5a3d4cb1]
[bt] (2) /usr/lib/x86_64-linux-gnu/libffi.so.6(ffi_call_unix64+0x4c) [0x7f0d4c88ce40]
[bt] (3) /usr/lib/x86_64-linux-gnu/libffi.so.6(ffi_call+0x2eb) [0x7f0d4c88c8ab]
[bt] (4) /usr/lib/python2.7/lib-dynload/_ctypes.x86_64-linux-gnu.so(_ctypes_callproc+0x48f) [0x7f0d4ca9c3df]
[bt] (5) /usr/lib/python2.7/lib-dynload/_ctypes.x86_64-linux-gnu.so(+0x11d82) [0x7f0d4caa0d82]
[bt] (6) /usr/bin/python2.7(PyObject_Call+0x43) [0x4b0cb3]
[bt] (7) /usr/bin/python2.7(PyEval_EvalFrameEx+0x5faf) [0x4c9faf]
[bt] (8) /usr/bin/python2.7(PyEval_EvalCodeEx+0x255) [0x4c2765]
[bt] (9) /usr/bin/python2.7(PyEval_EvalFrameEx+0x68d1) [0x4ca8d1]

Traceback (most recent call last):
File "./dff_rfcn/demo_batch.py", line 162, in
main()
File "./dff_rfcn/demo_batch.py", line 111, in main
arg_params=arg_params, aux_params=aux_params)
File "/home/huan/Deep-Feature-Flow/dff_rfcn/core/tester.py", line 30, in init
self._mod.bind(provide_data, provide_label, for_training=False)
File "/home/huan/Deep-Feature-Flow/dff_rfcn/core/module.py", line 839, in bind
for_training, inputs_need_grad, force_rebind=False, shared_module=None)
File "/home/huan/Deep-Feature-Flow/dff_rfcn/core/module.py", line 396, in bind
state_names=self._state_names)
File "/home/huan/Deep-Feature-Flow/dff_rfcn/core/DataParallelExecutorGroup.py", line 186, in init
self.bind_exec(data_shapes, label_shapes, shared_group)
File "/home/huan/Deep-Feature-Flow/dff_rfcn/core/DataParallelExecutorGroup.py", line 272, in bind_exec
shared_group))
File "/home/huan/Deep-Feature-Flow/dff_rfcn/core/DataParallelExecutorGroup.py", line 545, in _bind_ith_exec
context, self.logger)
File "/home/huan/Deep-Feature-Flow/dff_rfcn/core/DataParallelExecutorGroup.py", line 523, in _get_or_reshape
arg_arr = nd.zeros(arg_shape, context, dtype=arg_type)
File "/usr/local/lib/python2.7/dist-packages/mxnet-0.9.5-py2.7.egg/mxnet/ndarray.py", line 946, in zeros
return _internal._zeros(shape=shape, ctx=ctx, dtype=dtype)
File "/usr/local/lib/python2.7/dist-packages/mxnet-0.9.5-py2.7.egg/mxnet/_ctypes/ndarray.py", line 164, in generic_ndarray_function
c_array(ctypes.c_char_p, [c_str(val) for val in vals])))
File "/usr/local/lib/python2.7/dist-packages/mxnet-0.9.5-py2.7.egg/mxnet/base.py", line 78, in check_call
raise MXNetError(py_str(_LIB.MXGetLastError()))
mxnet.base.MXNetError: [10:46:34] src/c_api/c_api_ndarray.cc:390: Operator _zeros cannot be run; requires at least one of FCompute, NDArrayFunction, FCreateOperator be registered

Stack trace returned 10 entries:
[bt] (0) /usr/local/lib/python2.7/dist-packages/mxnet-0.9.5-py2.7.egg/mxnet/libmxnet.so(_ZN4dmlc15LogMessageFatalD1Ev+0x3c) [0x7f0d5982bf9c]
[bt] (1) /usr/local/lib/python2.7/dist-packages/mxnet-0.9.5-py2.7.egg/mxnet/libmxnet.so(MXImperativeInvoke+0x8d1) [0x7f0d5a3d4cb1]
[bt] (2) /usr/lib/x86_64-linux-gnu/libffi.so.6(ffi_call_unix64+0x4c) [0x7f0d4c88ce40]
[bt] (3) /usr/lib/x86_64-linux-gnu/libffi.so.6(ffi_call+0x2eb) [0x7f0d4c88c8ab]
[bt] (4) /usr/lib/python2.7/lib-dynload/_ctypes.x86_64-linux-gnu.so(_ctypes_callproc+0x48f) [0x7f0d4ca9c3df]
[bt] (5) /usr/lib/python2.7/lib-dynload/_ctypes.x86_64-linux-gnu.so(+0x11d82) [0x7f0d4caa0d82]
[bt] (6) /usr/bin/python2.7(PyObject_Call+0x43) [0x4b0cb3]
[bt] (7) /usr/bin/python2.7(PyEval_EvalFrameEx+0x5faf) [0x4c9faf]
[bt] (8) /usr/bin/python2.7(PyEval_EvalCodeEx+0x255) [0x4c2765]
[bt] (9) /usr/bin/python2.7(PyEval_EvalFrameEx+0x68d1) [0x4ca8d1]`

so is it the problem of mxnet install?

Hi,
I run dff_rfcn_test.py to test model's mAP.
However after it print the mAP, the program hang and never stop.
I find out there is multiprocessing.pool to do parallel evaluation. I think here cause the hang problem.
I hope there are some solution to exit program normally. Not kill -9 PID or os._exit(0).
Any idea how to fix it ?
Thanks very much !!

I'm trying to re-implement your work Deep-Feature-Flow on cityscapes dataset based on Deeplab(MXNet version).
Now, I have finished the training process, while I am confused how to evaluate on cityscapes dataset(which is not clear in your paper).

Should I use the sparse annotated frames or all frames of every clip?

If for the sparse annotated frames, does it means warpping its prev-frame's feature into this frame and compare the prediction with the label?

If for all frames, how to evaluate properly?

Thanks.

could anyone share the Imagenet VID dataset? the original webset http://vision.cs.unc.edu/ilsvrc2015/ui/vid seems down.

Hi, I was using the demo.py under ./rfcn & ./dff-rfcn folder. Then I checked the result in the ./demo/. The result of RFCN indicated that part worked fine. But for the part of DFF-RFCN, it could not reproduce the result of the demo video. Need I train this net myself to reproduce the result of the video?

Hi,
when I try to install mxnet using 'git checkout 62ecb60', I was faced with a error.
error: pathspec '62ecb60' did not match any file(s) known to git.
Could you please give me some advise?
Best wishes!

i ran this code on my pc.

but after installation of Anaconda2, i does not works.

several errors were occurred but i solved it(like GOMP4.0 or gcc errors)

and finally i ran this code on my ubuntu 16.04 with Anaconda2,

i met this error codes

name@name:~/Deep-Feature-Flow/rfcn$ python ./demo.py

Traceback (most recent call last): File "./demo.py", line 29, in <module> from core.tester import im_detect, Predictor File "/home/han/Deep-Feature-Flow/rfcn/core/tester.py", line 18, in <module> from nms.nms import py_nms_wrapper, cpu_nms_wrapper, gpu_nms_wrapper File "/home/han/Deep-Feature-Flow/rfcn/../lib/nms/nms.py", line 3, in <module> from cpu_nms import cpu_nms ImportError: /home/han/Deep-Feature-Flow/rfcn/../lib/nms/cpu_nms.so: undefined symbol: PyFPE_jbuf

what is this error code??? how could i fix it?

and there is another question.

what is difference between rfcn with dff_rfcn ???

can you please explain it??

because rfcn is better than dff_rfcn for my dataset.

I think that rfcn is not Deep-Feature-Flow one and dff_rfcn is applied Deep-Feature-Flow paper.

Is that right?

Answer me plz dear friends :)

hello, I run the dff demo(rfcn_dff) by following the instruction, but the demo result is not good as the show video. Some airplane could not be detected. Do you have some suggestions for this problem? Thanks very much.

Hi, @daijifeng001 @orpine,I use python interactive console to print mxnet.path,
it shows '/usr/local/lib/python2.7/dist-packages/mxnet-0.9.5-py2.7.egg/mxnet'

so I don't konw if it is right. I hope you can slove the AttributeError: 'module' object has no attribute 'MultiProposal'.

Thank you!

I want to visualize the network graph of training symbol of Flownet_RFCN. Kindly provide me proper inputs so that i can visualize the flow diagram. Kindly do the needful.