How to train my own model using my own data

[huayong]

Hi bittnt,

I had already installed crfasrnn and the code worked well.
But the code that you provided just contained the part using the trained model (TVG_CRFRNN_COCO_VOC.caffemodel) to get the result. How could I train a new model using my own data. Could you provide a pipeline about how to train a new model and the prototxt file used in training.
Any response will be appreciated.

Best wishes,
Huayong

[bittnt]

@huayong

We describe the training procedure in our ICCV 2015 paper (http://www.robots.ox.ac.uk/~szheng/papers/CRFasRNN.pdf). The parameters of the solver file for pascal voc dataset can be found in the paper. The prototxt and model are provided. The learning rate for each component of crf-rnn layer is given. To train the model on new dataset, you only need to set up your own solver file.

Thanks.

[mtourne]

It would be great if the code you used for it was provided as well :)

[cracker0dks]

👍 would be nice

[thuanvh]

I am trying to train my own images. I wonder how to create the training prototxt.
How do you use the label input? Which layer that you used for the loss calculation?

Thanks.

[thesby]

I want to train my own model based on this model. Any one can provide a train_val.prototxt? Thank you.

[sadeepj]

Closing old issues with no recent activity.