Hello. Thank you for uploading your great codes.
Recently, I'm working with these codes for semantic segmentation.
I made my config file based on deeplabv3plus model since this model provides config for iSAID and data I used is aerial & drone images. I finished setting my env and codes run pretty well.
But the problem is I couldn't find how to handle 'shuffle' pararmeter.
As I know, test.py returns output as a pickle format and it contains numpy array.
In my case, I used 84 png files(20002900) as a input and of course, output pickle file contains numpy array (20002900*84).
But when I unpickle to check the mask(np array), 84 arrays are in condtition of wrong order.
I expected results made in a same order as input data, but I couldn't find the way how to solve the problem.
ex) Input data : num101.png, num102.png ... num184.png
Output data(unpickled) : num123.png, num154.png ... num103.png
In test.py file, I could find data_loader has shuffle as False, but I couldn't understand why output pickle file contains np array in shuffled condition.
If I miss something, it would be thankful to let me know how to solve this problem.
Thank you.
Hello. Thank you for uploading your great codes.
Recently, I'm working with these codes for semantic segmentation.
I made my config file based on deeplabv3plus model since this model provides config for iSAID and data I used is aerial & drone images. I finished setting my env and codes run pretty well.
But the problem is I couldn't find how to handle 'shuffle' pararmeter.
As I know, test.py returns output as a pickle format and it contains numpy array.
In my case, I used 84 png files(20002900) as a input and of course, output pickle file contains numpy array (20002900*84).
But when I unpickle to check the mask(np array), 84 arrays are in condtition of wrong order.
I expected results made in a same order as input data, but I couldn't find the way how to solve the problem.
ex) Input data : num101.png, num102.png ... num184.png
Output data(unpickled) : num123.png, num154.png ... num103.png
In test.py file, I could find data_loader has shuffle as False, but I couldn't understand why output pickle file contains np array in shuffled condition.
If I miss something, it would be thankful to let me know how to solve this problem.
Thank you.