train on the base of the pretrained model SiamRPNotb.model

[zhanght021]

can this code train on the base of the pretrained model "SiamRPNotb.model" released by foolwood

[HelloRicky123]

yes，you can load the former 3 layers of the SiamRPNotb.model. The setting is in line 54 of the config.py file.

[zhanght021]

@HelloRicky123 Thanks for your code and your answer, and I have another questions

(1)in the 54version, you didn't freeze the former 3 BN layers, so is there any benefits? (2) if I change some code, can I load the full SiamRPNotb.model parameters, and then freeze the former 3 layers, and train on the base of the pretrained parameters .In my opinion ,SiamRPBotb.model is enough training, if I want to use in some special scene, can I base the fully model parameters, using some my own template to train (我可不可以改变一下几行代码，然后把SiamRPNotb.model的所有参数都导入，然后在固定住前三层，在这些已经训练好的参数上训练？因为我想着这个模型已经训练的很充分了，然后我用在比较特定的场景的话，可不可以在他整个模型参数的基础之上，用我自己的一些样本训练) Thank you very much

[fAnchao1]

@zhanght021 你用这个代码能够跑起来嘛，能指导我一下嘛

[zhanght021]

@fAnchao1 我还在下载数据集，等下完了试试，应该没问题的吧，作者都已经把结果测出来了

[fAnchao1]

可以加个qq？

[zhanght021]

@fAnchao1 这个等我明后天自己跑通了再交流，不然到时候就比较尴尬了，哈哈

[HelloRicky123]

@HelloRicky123 Thanks for your code and your answer, and I have another questions

(1)in the 54version, you didn't freeze the former 3 BN layers, so is there any benefits? (2) if I change some code, can I load the full SiamRPNotb.model parameters, and then freeze the former 3 layers, and train on the base of the pretrained parameters .In my opinion ,SiamRPBotb.model is enough training, if I want to use in some special scene, can I base the fully model parameters, using some my own template to train (我可不可以改变一下几行代码，然后把SiamRPNotb.model的所有参数都导入，然后在固定住前三层，在这些已经训练好的参数上训练？因为我想着这个模型已经训练的很充分了，然后我用在比较特定的场景的话，可不可以在他整个模型参数的基础之上，用我自己的一些样本训练) Thank you very much

Yes, of course. I tried this previously and can get 0.56 auc with code.

[fAnchao1]

@HelloRicky123 hello,how can i test the result in the vot-toolkit

[ghost]

@HelloRicky123 Thanks for your amazing work. And how to check the results with the vot-toolkit?

[fAnchao1]

@zhanght021 你做的怎么样了

[fAnchao1]

@longqiwf have you test on vot?

[stillqueen]

@fAnchao1 你现在问题解决了吗？要不要加个QQ？

[fAnchao1]

@stillqueen 还没有呢，你呢，可以啊加个qq我加你

[stillqueen]

@fAnchao1 974765289

[stillqueen]

@fAnchao1 上面是我QQ,你可以加一下哈

[HelloRicky123]

Closed first.