py-faster-rcnn 的常见错误

(1)缺少opencv
faster rcnn  ImportError: No module named cv2
=> conda install opencv

(2)参数错误

Traceback (most recent call last):
File "./tools/train_net.py", line 88, in <module>
cfg_from_list(args.set_cfgs)
File "/home/lab309/Documents/py-faster-rcnn/tools/../lib/fast_rcnn/config.py", line 269, in cfg_from_list
assert len(cfg_list) % 2 == 0
AssertionError

==>"[400, 500, 600, 700]" 参数，把中间的空格去掉变成"[400,500,600,700]"

(3)格式

Traceback (most recent call last):
  File "./tools/train_net.py", line 112, in <module>
    max_iters=args.max_iters)
  File "/home/yang/deep/Demo/Demo_D_R/py-faster-rcnn/tools/../lib/fast_rcnn/train.py", line 157, in train_net
    pretrained_model=pretrained_model)
  File "/home/yang/deep/Demo/Demo_D_R/py-faster-rcnn/tools/../lib/fast_rcnn/train.py", line 51, in __init__
    pb2.text_format.Merge(f.read(), self.solver_param)
AttributeError: 'module' object has no attribute 'text_format'

==> 在train.py加上
import google.protobuf.text_format

(4)numpy版本

Traceback (most recent call last):

  File "./tools/train_net.py", line 113, in <module>
    max_iters=args.max_iters)
  File "/home/yang/deep/Demo/Demo_D_R/py-faster-rcnn/tools/../lib/fast_rcnn/train.py", line 160, in train_net
    model_paths = sw.train_model(max_iters)
  File "/home/yang/deep/Demo/Demo_D_R/py-faster-rcnn/tools/../lib/fast_rcnn/train.py", line 101, in train_model
    self.solver.step(1)
  File "/home/yang/deep/Demo/Demo_D_R/py-faster-rcnn/tools/../lib/rpn/proposal_target_layer.py", line 66, in forward
    rois_per_image, self._num_classes)
  File "/home/yang/deep/Demo/Demo_D_R/py-faster-rcnn/tools/../lib/rpn/proposal_target_layer.py", line 191, in _sample_rois
    _get_bbox_regression_labels(bbox_target_data, num_classes)
  File "/home/yang/deep/Demo/Demo_D_R/py-faster-rcnn/tools/../lib/rpn/proposal_target_layer.py", line 127, in _get_bbox_regression_labels
    bbox_targets[ind, start:end] = bbox_target_data[ind, 1:]
TypeError: slice indices must be integers or None or have an __index__ method


==>
这里还是因为numpy版本的原因，最好的解决办法还是换numpy版本#conda list 查看numpy版本
pip install -U numpy==1.11.0 


(5)F0929 20:12:41.752293  2401 roi_pooling_layer.cu:91] Check failed: error == cudaSuccess (8 vs. 0)  invalid device function


==>
相关的问题https://github.com/rbgirshick/py-faster-rcnn/issues/2 
一般情况下都是因为显卡的计算能力不同而导致的，修改 py-faster-rcnn/lib/setup.py 的第135行，将arch改为与你显卡相匹配的数值
（比如我的Quadro K4000，计算能力是3.0，就将sm_35改成了sm_30）然后删除utils/bbox.c，nms/cpu_nms.c ，nms/gpu_nms.cpp ，在lib/下重新编译即可