Open cty-xiaobai opened 4 years ago
@cty-xiaobai Could you post the screenshot when AssertionError happens?
@cty-xiaobai Could you post the screenshot when AssertionError happens?
@tianzhi0549 sure
@cty-xiaobai Could you post the screenshot when AssertionError happens?
@tianzhi0549 Sorry to disturb you, do you have a good solution to my problem? thank you very much!
@cty-xiaobai Hi, I guess this error might be because you are using a wrong detectron2 version. Please try to use the detectron2 with commit id 9eb4831.
@cty-xiaobai Hi, I guess this error might be because you are using a wrong detectron2 version. Please try to use the detectron2 with commit id 9eb4831.
Ok i will try it, thanks again
@tianzhi0549 Can you provide a correct version of detectron2? I still have problems with the latest version, thank you!
@tianzhi0549 Can you provide a correct version of detectron2? I still have problems with the latest version, thank you!
@tianzhi0549 I have found the corresponding version, sorry to disturb you
The Condinst training set is 1080 x 1920 and 480 x 640 images, and verified on the 1080 x 1920 data set
the following error appeared AssertionError: ResizeShortestEdge(short_edge_length=(640, 672, 704, 736, 768, 800), max_size=1333, sample_style='choice')
So, I rewrote the mapper method so that although it can be trained, the results of the training will be worse (continue training on the basis of CondInst_R_101_BiFPN_3x_sem.pth)
This is the log file: log.txt
How should I modify the code to make my finetuning result better, not worse. Thank you for your help!