Py faster rcnn trainig error

After I downloaded VOCdevkit2007 data , the network trains perfectly(top picture). When it trains my data, it appears error(botton picture).

Hi,

Based on the log, the tool is searching for the JPEG image.
Do you use .jpg to store your database?

For other data format, please add the support here:

Thanks.

Yes , I store dataset used .jpeg.Screenshot from 2020-05-05 15-49-35

The two way have be right,but it has problem.

Hi,

The error indicates the the path doesn’t exist.
Could you check if the folder name is correct?

/home/bmw/…/JPEGImages/

Thanks.

My jpeg folder path is:" /home/bmw/py-faster-rcnn/data/VOCdevkit2007/VOC2007/JPEGImages "

I think this problem is train.txt,test.txt,val.txt and trainval.txt index’ wrongs. How to train my own data on pyfaster rcnn on?

Hi,

Sorry for the late update.

The error occurs from the index variable.
Suppose you should get the image patch like this:

    def image_path_from_index(self, index):
        ...
        image_path = os.path.join(self._data_path, 'JPEGImages',
                                  index + self._image_ext)

But the log shows the following:

/home/bmw/py-faster-rcnn/data/VOCdevkit2007/VOC2007/JPEGImages/.jpg

indicating the index value is empty.
Please help to check if any configure or file related to the index is missing first.

Thanks.