How to train with mobilenetv2

[Hanson0910]

when train with mobilenetv2,the errors occurred as following: Traceback (most recent call last): File "/home/fenghan/SNIPER/main_train.py", line 91, in sym = sym_inst.get_symbol_rpn(config) if config.TRAIN.ONLY_PROPOSAL else sym_inst.get_symbol_rcnn(config) File "/home/fenghan/SNIPER/symbols/faster/mobilenetv2_e2e.py", line 254, in get_symbol_rcnn name='multi_proposal_target') File "", line 107, in MultiProposalTarget File "SNIPER-mxnet/python/mxnet/_ctypes/symbol.py", line 135, in _symbol_creator s._compose(name=name, **kwargs) File "SNIPER-mxnet/python/mxnet/symbol/symbol.py", line 481, in _compose self.handle, name, num_args, keys, args)) File "SNIPER-mxnet/python/mxnet/base.py", line 149, in check_call raise MXNetError(py_str(_LIB.MXGetLastError())) mxnet.base.MXNetError: [11:34:07] src/core/symbolic.cc:72: Symbol.ComposeKeyword argument name crowd_boxes not found.

[chituma110]

I had the same problem ,did you solved the problem? @Hanson0910

[Andrea-Chang]

hi @Hanson0910 , Would you kindly provide the modified pre-trained model, or could you give me some hint about how to modify the model parameters?

I have tried:

1. download the pre-trained model from here: https://github.com/KeyKy/mobilenet-mxnet
2. then just put it in the related folder, and modified this line in config file pretrained: "./data/pretrained_model/mobilenet_v2"

very thanks.

[Andrea-Chang]

Hi @Hanson0910, Thanks for your reply. I downloaded the pre-trained model you provided above. I am a mxnet newbie, could you kindly give me some hints about how to modify the "class MobilenetV2" in MobileNetV2.py to fit the code parameters? Or should I do something else not just modify this file?

very thanks

[Andrea-Chang]

Hi, @Hanson0910 , Thanks for you reply. I load .params file and success get the dictionary. Should I find related layer name in class MobilenetV2 or class mobilenetv2_e2e?

Could you kindly show the example how to find the related layer name in class MobilenetV2(or mobilenetv2_e2e)?

For example: the key in xxxx.params : arg:conv1_bn_beta arg:conv1_bn_gamma arg:conv1_weight arg:conv2_1_dwise_bn_beta ... how can I find the related layer name in class MobilenetV2(or mobilenetv2_e2e)?

Thank you for your patience in answering all my questions.

[Hanson0910]

Give me your email.i send you the file. On 5/6/2019 16:17，Andrea-Changnotifications@github.com wrote：

Hi, @Hanson0910 , Thanks for you reply. I load .params file and success get the dictionary. Should I find related layer name in class MobilenetV2 or class mobilenetv2_e2e?

Could you kindly show the example how to find the related layer name in class MobilenetV2(or mobilenetv2_e2e)?

For example: the key in xxxx.params : arg:conv1_bn_beta arg:conv1_bn_gamma arg:conv1_weight arg:conv2_1_dwise_bn_beta ... how can I find the related layer name in class MobilenetV2(or mobilenetv2_e2e)?

Thank you for your patience in answering all my questions.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.

[Andrea-Chang]

hi, @Hanson0910, My email:

Thank you very much

[mehrazi]

@Hanson0910 can you send file for me? mehrdad.azizi72@gmail.com thanks in advanced

[yanliv]

@Hanson0910 I had the same problem, can you send file for me? yanli12321@qq.com Thank you very much

[LP940708]

@Hanson0910 I had the same problem, can you send file for me? 1220755365@qq.com Thank you very much

[czy1747]

@Hanson0910 I had the same problem, can you send file (mobilenetv2 and resnet50 pretained model) for me? 446115786@qq.com Thank you very much

[TianMingChen]

Give me your email.i send you the file. On 5/6/2019 16:17，Andrea-Changnotifications@github.com wrote： Hi, @Hanson0910 , Thanks for you reply. I load .params file and success get the dictionary. Should I find related layer name in class MobilenetV2 or class mobilenetv2_e2e? Could you kindly show the example how to find the related layer name in class MobilenetV2(or mobilenetv2_e2e)? For example: the key in xxxx.params : arg:conv1_bn_beta arg:conv1_bn_gamma arg:conv1_weight arg:conv2_1_dwise_bn_beta ... how can I find the related layer name in class MobilenetV2(or mobilenetv2_e2e)? Thank you for your patience in answering all my questions. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.

I had the same problem, can you send file for me? 2561363477@qq.com Thank you very much