“Domain Adaptive Faster R-CNN for Object Detection in the Wild“代码笔记

代码链接：https://github.com/krumo/Domain-Adaptive-Faster-RCNN-PyTorch
1、各种环境问题，重新建个环境，python3.6，pytorch版本1.0

2、CentOS和ubuntu 帖子看串了，没有yum

3、error: command ‘gcc’ failed with exit status 1的问题
https://blog.csdn.net/mm1327/article/details/106878984
搜错误技巧：搜具体error（从第一个开始），加上跑的工程名称

4、数据集的转换
Cityscapes–>CoCo形式
看代码，数据集输入错误，没有json文件，转换不了。

5、重新换个代码
https://github.com/tiancity-NJU/da-faster-rcnn-PyTorch
Q1：测试时，代码报错

CUDA Error: invalid argument, at line 147
CUDA Error: invalid argument, at line 154
Traceback (most recent call last):
  File "eval/test.py", line 271, in <module>
    rois_label = fasterRCNN(im_data, im_info, gt_boxes, num_boxes)
  File "/root/anaconda3/envs/DA-NJ/lib/python3.6/site-packages/torch/nn/modules/module.py", line 491, in __call__
    result = self.forward(*input, **kwargs)
  File "/root/Workspace/da-faster-rcnn-PyTorch-master/lib/model/faster_rcnn/faster_rcnn.py", line 51, in forward
    rois, rpn_loss_cls, rpn_loss_bbox = self.RCNN_rpn(base_feat, im_info, gt_boxes, num_boxes)
  File "/root/anaconda3/envs/DA-NJ/lib/python3.6/site-packages/torch/nn/modules/module.py", line 491, in __call__
    result = self.forward(*input, **kwargs)
  File "/root/Workspace/da-faster-rcnn-PyTorch-master/lib/model/rpn/rpn.py", line 78, in forward
    im_info, cfg_key))
  File "/root/anaconda3/envs/DA-NJ/lib/python3.6/site-packages/torch/nn/modules/module.py", line 491, in __call__
    result = self.forward(*input, **kwargs)
  File "/root/Workspace/da-faster-rcnn-PyTorch-master/lib/model/rpn/proposal_layer.py", line 159, in forward
    output[i,:num_proposal,1:] = proposals_single
RuntimeError: dimension out of range (expected to be in range of [-1, 0], but got 1)

不知道如何下手：找termina的上面报出更多的信息

(DA-NJ) root@310566561f10:~/Workspace/da-faster-rcnn-PyTorch-master# CUDA_VISIBLE_DEVICES=0 python eval/test.py --dataset cityscape --part test_t --model_dir=/root/Workspace/da-faster-rcnn-PyTorch-
master/data/ztc/adaptation/Experiment/da_model/vgg16/cityscape/cityscape_consist_default.pth
eval/..
Called with args:
Namespace(cfg_file='cfgs/vgg16.yml', checkepoch=1, checkpoint=10021, checksession=1, class_agnostic=False, cuda=False, dataset='cityscape', large_scale=False, mGPUs=False, model_dir='/root/Workspac
e/da-faster-rcnn-PyTorch-master/data/ztc/adaptation/Experiment/da_model/vgg16/cityscape/cityscape_consist_default.pth', model_name='res101.bs1.pth', net='vgg16', parallel_type=0, part='test_t', set
_cfgs=None, vis=False)
WARNING: You have a CUDA device, so you should probably run with --cuda    **！！！！！！！！！！！！！！！！！！！！！！！！！！！！！**
loading our dataset...........
/root/Workspace/da-faster-rcnn-PyTorch-master/lib/model/utils/config.py:376: YAMLLoadWarning: calling yaml.load() without Loader=... is deprecated, as the default Loader is unsafe. Please read http
s://msg.pyyaml.org/load for full details.
  yaml_cfg = edict(yaml.load(f))
Using config:

提示：没有用到cuda
解决方法：
1、在cfg里改配置文件，此处cfg里没有该选项
2、在运行的这个代码里（test.py）文件里修改，本来是：

parser.add_argument('--cuda', dest='cuda',
                      help='whether use CUDA',
                      action='store_true')

修改为：

parser.add_argument('--cuda', default='True',dest='cuda',
                      help='whether use CUDA',
                      action='store_true')

扩展阅读：该parser里的dest：（parser教程）
https://blog.csdn.net/u014397533/article/details/51075824

此代码修改的地方：（方便以后学习）
一些路径，有的改成了绝对路径。。等

6、迷惑之处：
cityscapes数据集 foggy的数量到底多少，哪里下载，为啥有的只有500张，我下载的25G。。。
foggy-cityscapes到底怎么转成coco形式？
cityscapescripts怎么用？
代码进一步debug？
batch_demo用法？