training on own dataset

davidsandberg / facenet

Face recognition using Tensorflow

MIT License

13.85k stars 4.81k forks source link

training on own dataset #678

Closed PythonImageDeveloper closed 6 years ago

PythonImageDeveloper commented 6 years ago

How i train this model on own dataset ? is it possible for training ? my dataset is breeds of dogs.

huangbiubiu commented 6 years ago

Absolutely you can. These are parameters you should pay attention:

--data_dir: you own training dir with the same structure as LFW
--lfw_pairs and --lfw_dir for testing.

Then you can train this without making any changes to the code.

PythonImageDeveloper commented 6 years ago

Thanks Please put the step by step for training and evaluation. i don't find any python script file for training? and i my own dataset is not face but these are fingerprint , my question is that : is that codes possible for training for that dataset ?

huangbiubiu commented 6 years ago

@zeynali Yes you can train this for any type data. Just run train_tripletloss.py with required parameters

erdalalim commented 5 years ago

hi mr. @huangbiubiu , I want train facenet model with Casia Web Face dataset, and then to test it on LFW dataset.Its possible? I will try to do augmentation techniques on Casia Web dataset, If I will do good augmentation on the data it will give better accuracy?

huangbiubiu commented 5 years ago

@erdalalim

Your first question. Yes. Certainly, it's possible.

Your second question, yes. If the augmentation technic is good it will definitely give a good result on the testing set.

erdalalim commented 5 years ago

@huangbiubiu thanks for reply. You Run facenet before? I wanna run facenet according to :https://github.com/TessFerrandez/research-papers/blob/master/facenet/FaceNet.ipynb because there is video explanation : https://www.youtube.com/watch?v=w--c0qG9MCc

the link with .ipynb is the same one like the code in this repository?