Is it possible to create a text-image model from scratch, based on a database fed by the creations of artists who have consented to it, to prove that it’s possible?
This is my diploma project.
My deadline is in two weeks.
I’m a student.
I have little or no money to hire someone to program this for me.
Modified by Moderator
Can I do it myself?
Should I use TensorFlow, Craiyon, something else?
Thanks for the little guidance !
Is possible to finetune a pretrained model that will recognize that images. the issue here is that you will do a clasification in 2 class instead of many and that you will have to add the images with class 0 = NoRights to use and class 1 = rights to use
Then you feed many images equal if possible to the model for the 2 class and finetune the model to give u this 2 class. Issue is that you must literally retrain every time an artist join.
Model you get a resnet easy and finetune it for your need.
Plently of free colab around.
finetuned-resnet50
or
finetuning