let x_train=(a list of images), let y_train=(a list of labels like [[age, gender], ..., [age, gender]]). Assuming you have a fairly balanced training set among age and gender, you can split them into a training and validation set. If using TF's sequential class, check the fit() function's arguments for where to do this. As for splitting the files into different directories, do you really need to do that?
#Splitting up data
static cedar
I'd create a structure like
/dataset/train/
/dataset/validation/
and if the dataet you downloaded doesn't already split it for you, TF lets you do that in model.fit()
floral joltBOT
@split pivot has leveled up! (0 ➜ 1)
static cedar
When you train a model, it accepts inputs and labels as tensors (or numpy arrays). That means the images must be converted to tensors. For this you can search online for guides on how to do that, such as https://www.tensorflow.org/tutorials/load_data/images#load_data_using_a_keras_utility
which details how you can create a tf.data.Dataset object that can be fed into the model
static cedar
I don't know how to interpret that vertical blue line but everything else seems normal, loss goes down and accuracy goes up
for here it's just tweaking the hyperparameters to get the best result
static cedar
Check for class balance and that your labels are correctly made. It wouldn't make sense to validate on a set od 99 dogs and 1 cat.