training loss

[caoyangcr7]

Thanks for your work firstly. when I train my own dataset---10 types of bmw cars, totally 512 images (the url is http://ai.stanford.edu/~jkrause/cars/car_dataset.html), but the loss nearly can't be cut down when I using your original code or I changed the code like https://github.com/akarshzingade/image-similarity-deep-ranking/issues/17. That's confused... Anyone meets the same issue with me ? Thanks for your reply.