I am working on a network implementation where I start from an already trained network model (AlexNet) on imagenet and I train it with my own dataset.
I wanted to know if there is a way to print the classes the network has been trained on previously ?
And also can someone explain how the network keep track of the classes it has been trained on to retrieve them when it is tested with a image for example ?
Thank you in advance
Based on what you’ve said, I think you misinterpret what the network “learn” on training (sorry if I am wrong). The network doesn’t know anything specific about the class (name or something); it only know that a picture may belong to the i-th class, where the order of classes is decided when training the network.
If it’s trained on ImageNet, you can find the ground truth by visiting ImageNet’s website. Also, you can “test” it (if it has acceptable accuracy) by feeding the dataset it is pretrained on, so you can get the order of the classes.
Thank you for the answer, in fact I did’nt express myself correctly. And I understood the stupidity of my question after posting it.
Sorry for this.