Pytorch_Retinaface
Pytorch_Retinaface copied to clipboard
biubug6
evaluation.py error
hi, thanks for you share. First i run this command: python setup.py build_ext --inplace and then i run: python evaluation.py -p widerface_txt/ -g ground_truth/ i get error like this: ImportError: dynamic module does not define module export function (PyInit_bbox) , it seems that i have not succeed in compile box_overlaps.pyx. but what should i do. my env is python3.5 and pytorch1.2.0.
Sorry, I didn't come across this problem. My env is python3.7 and pytorch1.2.0 ,ubuntu18.04 You can refer to https://github.com/wondervictor/WiderFace-Evaluation.
@wanglaotou Hi, I met the same question, have you solved this problem?
Hello, I think I found the answer.
The problem occurred due to name mismatch.
You need to change the output so file name in setup.py:
change
package = Extension('bbox', ['box_overlaps.pyx'], include_dirs=[numpy.get_include()])
to
package = Extension('box_overlaps', ['box_overlaps.pyx'], include_dirs=[numpy.get_include()])
and change module name in evaluation.py to match the names:
from box_overlaps import bbox_overlaps
@yechanp Thanks for your answer, I had finished this issues. I just replaced cpython version bbox.pyx with python version bbox.py.
@BlueCCircles Hello, Thanks for your answer. Could you share your python version bbox.py, please?
使用如下的代码去替代 evaluation.py中的 bbox_overlaps，并注释掉from box_overlaps import bbox_overlaps，并且可以不使用 python setup.py build_ext --inplace这行命令； python代码如下： def bbox_overlaps_(bboxes1, bboxes2): N = bboxes1.shape[0] M = bboxes2.shape[0] overlaps = np.zeros((N, M))
for i in range(N):
bbox1 = bboxes1[i]
area1 = (bbox1[2] - bbox1[0]) * (bbox1[3] - bbox1[1])
for j in range(M):
bbox2 = bboxes2[j]
area2 = (bbox2[2] - bbox2[0]) * (bbox2[3] - bbox2[1])
x_min = max(bbox1[0], bbox2[0])
y_min = max(bbox1[1], bbox2[1])
x_max = min(bbox1[2], bbox2[2])
y_max = min(bbox1[3], bbox2[3])
intersection_area = max(0, x_max - x_min) * max(0, y_max - y_min)
union_area = area1 + area2 - intersection_area
overlaps[i, j] = intersection_area / union_area
return overlaps
