Closed nandini211995 closed 3 years ago
set USE_GN=False in cfgs.py
Dear authors,
I am trying to perform testing of the DOTA dataset on your provided benchmark model. I followed all the steps which you mentioned on the Github page ,but I am not able to get the result. Here I am mentioning all the steps which I followed. Kindly guide me to obtain testing results on DOTA images.
Kindly guide me for the same . It will be very helpful .
Thank you
Nandini Saini IIT Jodhpur
On Thu, Jun 10, 2021 at 9:50 AM yangxue @.***> wrote:
Closed #18 https://github.com/yangxue0827/RotationDetection/issues/18.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/yangxue0827/RotationDetection/issues/18#event-4868820092, or unsubscribe https://github.com/notifications/unsubscribe-auth/APYOHQ2UWXNQXLFHDEAHSU3TSA4RTANCNFSM46JM4SRQ .
A gentle reminder for the request which is below written.
I am trying to perform testing of the DOTA dataset on your provided benchmark model. I followed all the steps which you mentioned on the Github page ,but I am not able to get the result. Here I am mentioning all the steps which I followed. Kindly guide me to obtain testing results on DOTA images.
- Install all dependencies i.e. python version : 3.7.10 , tensorflow version : 1.15.4
- Download the pretrained weights and here i am attaching the image of directory of weights
- Follow all the steps of compile and testing
Kindly guide me for the same . It will be very helpful .
Thank you
Nandini Saini IIT Jodhpur
On Thu, Jun 10, 2021 at 9:50 AM yangxue @.***> wrote:
Closed #18 https://github.com/yangxue0827/RotationDetection/issues/18.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/yangxue0827/RotationDetection/issues/18#event-4868820092, or unsubscribe https://github.com/notifications/unsubscribe-auth/APYOHQ2UWXNQXLFHDEAHSU3TSA4RTANCNFSM46JM4SRQ .
You need to train a model based on pretrain model before testing. If you just want to use my trained model, please download the trained weight according to the table in readme, and then put the weights in the right path (output/trained_weights/).
Dear authors,
I just used your trained model provides in this link https://github.com/DetectionTeamUCAS/Models/tree/master/R2CNN-Plus-Plus_Tensorflow because I want to perform a test operation on DOTA test images. I put all the weights and trained weights in the correct folder and run it. It is running but in the output folder as you mentioned (the output file or visualization is in $PATH_ROOT/tools/#DETECTOR/test_dota/VERSION): there are images without any detection or any result. Kindly guide me to the next steps.
Thanks and Regards Nandini Saini Ph.D. Scholar IIT Jodhpur
On Mon, Jun 14, 2021 at 3:36 PM yangxue @.***> wrote:
You need to train a model based on pretrain model before testing. If you just want to use my trained model, please download the trained weight according to the table in readme, and then put the weights in the right path (output/trained_weights/).
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/yangxue0827/RotationDetection/issues/18#issuecomment-860564137, or unsubscribe https://github.com/notifications/unsubscribe-auth/APYOHQ47GQQJ3AOS3ONZVB3TSXICHANCNFSM46JM4SRQ .
While running testing on DOTA images, this error is occuring. AttributeError: module 'libs.configs.cfgs' has no attribute 'USE_GN'