InvalidArgumentError with dataset from CSVReader

[clicit]

I successfully created a custom dataset with the CSVReader. But when I want to use it for local training with default FRCNN config, I get the following error:

Traceback (most recent call last):
  File "c:\path\to\python\python36\lib\site-packages\tensorflow\python\client\session.py", line 1323, in _do_call
    return fn(*args)
  File "c:\path\to\python36\lib\site-packages\tensorflow\python\client\session.py", line 1302, in _run_fn
    status, run_metadata)
  File "c:\path\to\python\python36\lib\site-packages\tensorflow\python\framework\errors_impl.py", line 473, in __exit__
    c_api.TF_GetCode(self.status.status))
tensorflow.python.framework.errors_impl.InvalidArgumentError: Nan in summary histogram for: fasterrcnn/rpn/rpn_bbox_target/histogram
         [[Node: fasterrcnn/rpn/rpn_bbox_target/histogram = HistogramSummary[T=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](fasterrcnn/rpn/rpn_bbox_target/histogram/tag, fasterrcnn/rpn/anchor_target/ScatterNd)]]

The images all have a resolution of 640x480 and are labeled with class "car". I tried the suggestions from issue #133 , without success.

Any ideas? Thanks!

[vierja]

Hi @clicit,

Does it fail right away or after a while?

[clicit]

Hi @vierja , it fails after some minutes before any images are processed. Thanks

Am 16. Januar 2018 17:20:16 MEZ schrieb Javier Rey notifications@github.com:

Hi @clicit,

Does it fail right away or after a while?

-- You are receiving this because you were mentioned. Reply to this email directly or view it on GitHub: https://github.com/tryolabs/luminoth/issues/149#issuecomment-358016980

-- Diese Nachricht wurde von meinem Android-Mobiltelefon mit K-9 Mail gesendet.