search

philkr / voc-classification

Train and evaluate VOC 2007 / 2012 image classification models. We used this code to evaluate several feature learning algorithms.
BSD 2-Clause "Simplified" License
48 stars 22 forks source link

VOC Classification Code is not working for my model that has the alexnet structure #6

Open Maryames opened 6 years ago

Maryames commented 6 years ago

Hi,

I have a pretrained model in torch and I have transfered the weights and biases to the caffe, and I checked if the model is transfered correctly. The problem is that when I want to evaluate my model for pascal dataset with philkr VOC classification code (https://github.com/philkr/voc-classification), it gives me 7% and it seems that it cannot finetune. I want to use the code with clipping from drop6 instead of drop7 which is the default in the code and I changed the code accordingly, but I still get the same result.I was wondering how I can use the code correctly?I really appreciate any help

Thank you so much,

uahsan3 commented 6 years ago

Hello,

I am facing the same problem. I am trying to finetune my model on Pascal VOC 2007 dataset and the loss is not going down beyond random. It's not clear what is going wrong.

I have tried with default hyperparameter settings as well as my own.

Please advise.

Thanks

jianghaojun commented 5 years ago

I encountered the same problem as @Maryames . And I solved it with the help of @uahsan3 . If anyone else encounter this problem, please look at the trainval.prototxt generate by the code (in train_cls.py)and check the data layer. The code will create a right data layer and if your alexnet.prototxt also include a data layer, then the network will have two data layer and the image data will not go through the network. Besides, I attach two screenshots of trainval.prototxt below, the first is wrong and the second is right. bug1

bug2