ModuleNotFoundError: No module named 'lib.fpn.box_intersections_cpu.bbox'

[PrimoWW]

thanks for your great job.
when I followed README, and try to run the train command:

python train.py -mode predcls -datasize large -data_path $DATAPATH

I had some problem.
First, in dataloader/action_genome.py, from scipy.misc import imread . the lib scipy seems not surport the fuction imread in latest version. so i change it to from imageio import imread

Second, when I continue totry to run the project, a new problem occured:
ModuleNotFoundError: No module named 'lib.fpn.box_intersections_cpu.bbox'
it means i didnt generate the file in this directory, so I went back to

cd fpn/box_intersections_cpu
python setup.py install ( i found if i dont add 'install ' , the command cant work)

and then

~/project/STTran/lib/fpn/box_intersections_cpu$ python setup.py install
running install
running build
running build_ext
running install_lib
running install_egg_info
Removing /home/liujingwei/anaconda3/envs/scene_graph_benchmark/lib/python3.7/site-packages/bbox_cython-0.0.0-py3.7.egg-info
Writing /home/liujingwei/anaconda3/envs/scene_graph_benchmark/lib/python3.7/site-packages/bbox_cython-0.0.0-py3.7.egg-info

So, it didnt generate the file(lib.fpn.box_intersections_cpu.bbox), and i dont know how to fix it, then I may need your help.^_^

[yrcong]

1. My scipy version is 1.1.0. You could switch to this version.
2. You could use

python setup.py build_ext --inplace

to compile the box_intersections_cpu/draw_rectangles code.

Please let me know if you still meet some problems:)

[PrimoWW]

thanks for your advice! when I fixed these problem with your help. Then I countinued to run the code, and the program raise a ERROR

ImportError: cannot import name '_C' from 'fasterRCNN.lib.model' (/home/xxx/project/STTran/fasterRCNN/lib/model/init.py)

so ,I solved the problem by running (maybe you should add this command into README)

/STTran/fasterRCNN/lib$ python setup.py build_ext --inplace

and then , I met this error:

Traceback (most recent call last):
File "train.py", line 41, in
object_detector = detector(train=True, object_classes=AG_dataset_train.object_classes, use_SUPPLY=True, mode=conf.mode).to(device=gpu_device)
File "/home/xxx/project/STTran/lib/object_detector.py", line 29, in init
checkpoint = torch.load('fasterRCNN/models/faster_rcnn_ag.pth.pth')
File "/home/xxx/anaconda3/envs/scene_graph_benchmark/lib/python3.7/site-packages/torch/serialization.py", line 381, in load
f = open(f, 'rb')
FileNotFoundError: [Errno 2] No such file or directory: 'fasterRCNN/models/faster_rcnn_ag.pth.pth'

I found the reason is that In the file lib/object_detector.py , 29th line,
checkpoint = torch.load('fasterRCNN/models/faster_rcnn_ag.pth.pth')
it seems that there are double 'pth'.

when I solved these problem , I can run the program successfully. but unfortunately I found there is not enough GPU memory that I need. So I may continue to ask questions when I have enough GPU memory. ：）

[yrcong]

Thanks for reminding! There is a redundant suffix exactly.

For GPU, I used RTX2080ti (ca. 11GB memory).

[PrimoWW]

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
loading word vectors from data/glove.6B.200d.pt
loading word vectors from /home/xxx/Dokumente/neural-motifs-master/data/glove.6B.200d.pt
background -> background
fail on background
Traceback (most recent call last):
File "train.py", line 102, in
entry = object_detector(im_data, im_info, gt_boxes, num_boxes, gt_annotation ,im_all=None)
File "/home/xxx/anaconda3/envs/scene_graph_benchmark/lib/python3.7/site-packages/torch/nn/modules/module.py", line 547, in call
result = self.forward(*input, **kwargs)
File "/home/xxx/project/STTran/lib/object_detector.py", line 306, in forward
FINAL_FEATURES = self.fasterRCNN.RCNN_roi_align(FINAL_BASE_FEATURES, FINAL_BBOXES)
File "/home/xxx/anaconda3/envs/scene_graph_benchmark/lib/python3.7/site-packages/torch/nn/modules/module.py", line 547, in call
result = self.forward(*input, kwargs)
File "/home/xxx/project/STTran/fasterRCNN/lib/model/roi_layers/roi_align.py", line 58, in forward
input, rois, self.output_size, self.spatial_scale, self.sampling_ratio
File "/home/xxx/project/STTran/fasterRCNN/lib/model/roi_layers/roi_align.py", line 20, in forward
output = _C.roi_align_forward(input, roi, spatial_scale, output_size[0], output_size[1], sampling_ratio)
RuntimeError: Not compiled with GPU support (ROIAlign_forward at /home/xxx/project/STTran/fasterRCNN/lib/model/csrc/ROIAlign.h:21)
frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits, std::allocator > const&) + 0x6d (0x7fc22b3171cd in /home/xxx/anaconda3/envs/scene_graph_benchmark/lib/python3.7/site-packages/torch/lib/libc10.so)
frame #1: + 0xb314 (0x7fc1f4145314 in /home/xxx/project/STTran/fasterRCNN/lib/model/_C.cpython-37m-x86_64-linux-gnu.so)
frame #2: + 0x22164 (0x7fc1f415c164 in /home/xxx/project/STTran/fasterRCNN/lib/model/_C.cpython-37m-x86_64-linux-gnu.so)
frame #3: + 0x1d7f4 (0x7fc1f41577f4 in /home/xxx/project/STTran/fasterRCNN/lib/model/_C.cpython-37m-x86_64-linux-gnu.so)

frame #8: THPFunction_apply(_object, _object) + 0xa56 (0x7fc24dbda5f6 in /home/xxx/anaconda3/envs/scene_graph_benchmark/lib/python3.7/site-packages/torch/lib/libtorch_python.so)
frame #41: __libc_start_main + 0xf3 (0x7fc25c8200b3 in /lib/x86_64-linux-gnu/libc.so.6)

there are a problem that i cant solve, i guess that maybe I cant use the C lib.

[PrimoWW]

thank you for your help! I ran the code successfully