Open zimonitrome opened 4 years ago
yes please I do not have enough cpu or storage to prepare dataset and make training..
pre-trained model would be very useful
Hi, @koren88i could you re-upload the pretrained weights please, i don't have the resources to train ttnet.
@koren88i Could you please provide the pretrained weights again? The link before was expired. Thanks a lot.
@koren88i Could please kindly help to provide the pre-trained model? We cannot follow the above link to get the file. Thanks a lot!
my trained model did not work... 0.6 ball bounce all the time. If I succeed I will upload again. Sorry for the delay.
got it. thanks!
在 2021年3月26日,17:19,koren88i @.***> 写道:
my trained model did not work... 0.6 ball bounce all the time. If I succeed I will upload again. Sorry for the delay.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.
my trained model did not work... 0.6 ball bounce all the time. If I succeed I will upload again. Sorry for the delay.
Did you succeed now? I also have similar problems. I implemented the same training strategy of the repo but I could not get accurate predictions.
Hi all,
Can I get pertained models?
Thanks
hi guys anyone get the petrained model?? please
my trained model did not work... 0.6 ball bounce all the time. If I succeed I will upload again. Sorry for the delay.
hi my friend ..can you help me how did you trained the model please ..the same strategy they mentioned in their project?
Could anyone provide pretrained weights please?
Do you have the pretrained weights?
Hi, can anyone share weights for me ?
Hello, I ran the training and testing with the settings available in the files and could not reproduce the results either. The segmentation of the players is OK, but not the ball detection. Most of time there is no detection, and if a detection occurs, the position is always around the middle of the image far away from ground truth. The problem occurs already at the Global stage (1st training).
Hello, I ran the training and testing with the settings available in the files and could not reproduce the results either. The segmentation of the players is OK, but not the ball detection. Most of time there is no detection, and if a detection occurs, the position is always around the middle of the image far away from ground truth. The problem occurs already at the Global stage (1st training).
i am facing the same issue?
my trained model did not work... 0.6 ball bounce all the time. If I succeed I will upload again. Sorry for the delay.
@koren88i Hi, have you succeeded?
It would be useful if pre-trained weights could be provided.
Are there any plans to making them available?