mitosis detection（二）

_LYH_

于 2020-03-01 20:50:22 发布

阅读量318

点赞数

分类专栏： mitosis caffe

本文链接：https://blog.csdn.net/qq_41574648/article/details/104599825

版权

mitosis 同时被 2 个专栏收录

8 篇文章 0 订阅

订阅专栏

caffe

8 篇文章 0 订阅

订阅专栏

I0301 12:36:30.536449 25955 solver.cpp:245]     Train net output #0: loss_bbox = 0.00237033 (* 1 = 0.00237033 loss)
I0301 12:36:30.536458 25955 solver.cpp:245]     Train net output #1: loss_cls = 0.0697297 (* 1 = 0.0697297 loss)
I0301 12:36:30.536464 25955 solver.cpp:245]     Train net output #2: rpn_cls_loss = 0.0507008 (* 1 = 0.0507008 loss)
I0301 12:36:30.536470 25955 solver.cpp:245]     Train net output #3: rpn_loss_bbox = 0.10025 (* 1 = 0.10025 loss)
I0301 12:36:30.536478 25955 sgd_solver.cpp:106] Iteration 260, lr = 0.001
I0301 12:36:34.376888 25955 solver.cpp:229] Iteration 280, loss = nan
I0301 12:36:34.376953 25955 solver.cpp:245]     Train net output #0: loss_bbox = 0 (* 1 = 0 loss)
I0301 12:36:34.376965 25955 solver.cpp:245]     Train net output #1: loss_cls = 0.116323 (* 1 = 0.116323 loss)
I0301 12:36:34.376971 25955 solver.cpp:245]     Train net output #2: rpn_cls_loss = 5.43175 (* 1 = 5.43175 loss)
I0301 12:36:34.376977 25955 solver.cpp:245]     Train net output #3: rpn_loss_bbox = nan (* 1 = nan loss)
I0301 12:36:34.376984 25955 sgd_solver.cpp:106] Iteration 280, lr = 0.001
./experiments/scripts/faster_rcnn_end2end_12.sh: line 53: 25955 Floating point exception(core dumped) ./tools/train_net.py --gpu ${GPU_ID} --solver models/${PT_DIR}/${NET}/faster_rcnn_end2end/solver.prototxt --weights data/imagenet_models/${NET}.v2.caffemodel --imdb ${TRAIN_IMDB} --iters ${ITERS} --cfg experiments/cfgs/faster_rcnn_end2end_2012.yml ${EXTRA_ARGS}

参考https://www.cnblogs.com/bile/p/9110727.html

经过分析调试，发现这个问题还是自己的数据集标注越界的问题！！！越界有6种形式：x1<0; x2>width; x2<x1; y1<0; y2>height; y2<y1。不巧的是，源代码作者是针对pascal_voc数据写的，压根就没有考虑标注出错的可能性。发布的代码中只在append_flipped_images函数里 assert (boxes[:, 2] >= boxes[:, 0]).all()，也就是只断言了水平翻转后的坐标x2>=x1，这个地方报错可能是x的标注错误，参考前面的错误2。但是，对于y的标注错误，根本没有检查。

Caffe-Faster RCNN错误：TypeError: ‘numpy.float64’ object cannot be interpreted as an index

参考https://blog.csdn.net/CAU_Ayao/article/details/84679340

在训练Faster RCNN时候出现以下错误：

Solving...
Process Process-3:
Traceback (most recent call last):
  File "/usr/lib/python2.7/multiprocessing/process.py", line 267, in _bootstrap
    self.run()
  File "/usr/lib/python2.7/multiprocessing/process.py", line 114, in run
    self._target(*self._args, **self._kwargs)
  File "./tools/train_faster_rcnn_alt_opt.py", line 197, in train_fast_rcnn
    max_iters=max_iters)
  File "/home/liguangyao/Programming/caffe-faster-rcnn/tools/../lib/fast_rcnn/train.py", line 161, in train_net
    model_paths = sw.train_model(max_iters)
  File "/home/liguangyao/Programming/caffe-faster-rcnn/tools/../lib/fast_rcnn/train.py", line 102, in train_model
    self.solver.step(1)
  File "/home/liguangyao/Programming/caffe-faster-rcnn/tools/../lib/roi_data_layer/layer.py", line 144, in forward
    blobs = self._get_next_minibatch()
  File "/home/liguangyao/Programming/caffe-faster-rcnn/tools/../lib/roi_data_layer/layer.py", line 63, in _get_next_minibatch
    return get_minibatch(minibatch_db, self._num_classes)
  File "/home/liguangyao/Programming/caffe-faster-rcnn/tools/../lib/roi_data_layer/minibatch.py", line 55, in get_minibatch
    num_classes)
  File "/home/liguangyao/Programming/caffe-faster-rcnn/tools/../lib/roi_data_layer/minibatch.py", line 100, in _sample_rois
    fg_inds, size=fg_rois_per_this_image, replace=False)
  File "mtrand.pyx", line 1192, in mtrand.RandomState.choice
TypeError: 'numpy.float64' object cannot be interpreted as an index

   
   
   
   1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23

在训练stage1 rpn时，出现’numpy.float64’ object cannot be interpreted as an index 的提示错误，几乎所有的博客中都指出，需要更换numpy 的版本，照做之后，出现ImportError: numpy.core.multiarray failed to import，这个问题又是numpy不匹配造成的，这样就形成了恶性循环，所以，可以考虑从根源上解决’numpy.float64’ object cannot be interpreted as an index的问题。
需要修改一下几个文件：

1. /home/xxx/py-faster-rcnn/lib/roi_data_layer/minibatch.py

将第26行：
fg_rois_per_image = np.round(cfg.TRAIN.FG_FRACTION * rois_per_image)
改为：
fg_rois_per_image = np.round(cfg.TRAIN.FG_FRACTION * rois_per_image).astype(np.int)

第174，175行改为：
for ind in inds:
cls = clss[ind]
start =int( 4 * cls) //改这句
end = int(start + 4) //改这句
bbox_targets[ind, start:end] = bbox_target_data[ind, 1:]
bbox_inside_weights[ind, start:end] = cfg.TRAIN.BBOX_INSIDE_WEIGHTS

2. /home/xxx/py-faster-rcnn/lib/datasets/ds_utils.py

将第12行：hashes = np.round(boxes * scale).dot(v)
改为：hashes = np.round(boxes * scale).dot(v).astype(np.int)

 
 
 
 1
2

3. /home/xxx/py-faster-rcnn/lib/fast_rcnn/test.py

将第129行： hashes = np.round(blobs['rois'] * cfg.DEDUP_BOXES).dot(v)
改为： hashes = np.round(blobs['rois'] * cfg.DEDUP_BOXES).dot(v).astype(np.int)

 
 
 
 1
2

4. /home/xxx/py-faster-rcnn/lib/rpn/proposal_target_layer.py

将第60行：fg_rois_per_image = np.round(cfg.TRAIN.FG_FRACTION * rois_per_image)
改为：fg_rois_per_image = np.round(cfg.TRAIN.FG_FRACTION * rois_per_image).astype(np.int)

 
 
 
 1
2

解决完上一个问题后，又出现 TypeError: slice indices must be integers or None or have an index method的问题，如果没有改变numpy的版本。

5. 修改 /home/XXX/py-faster-rcnn/lib/rpn/proposal_target_layer.py，转到123行：

for ind in inds:
        cls = clss[ind]
        start = 4 * cls
        end = start + 4
        bbox_targets[ind, start:end] = bbox_target_data[ind, 1:]
        bbox_inside_weights[ind, start:end] = cfg.TRAIN.BBOX_INSIDE_WEIGHTS
    return bbox_targets, bbox_inside_weights

 
 
 
 1
2
3
4
5
6
7

这里的ind，start，end都是 numpy.int 类型，这种类型的数据不能作为索引，所以必须对其进行强制类型转换，转化结果如下：

for ind in inds:
        ind = int(ind)
        cls = clss[ind]
        start = int(4 * cos)
        end = int(start + 4)
        bbox_targets[ind, start:end] = bbox_target_data[ind, 1:]
        bbox_inside_weights[ind, start:end] = cfg.TRAIN.BBOX_INSIDE_WEIGHTS
    return bbox_targets, bbox_inside_weight

 
 
 
 1
2
3
4
5
6
7
8