afhq dataset error

[ZYJ-JMF]

Hi,when I use your actual example in Train: python main.py --gpu 0 --dataset afhq_cat --output_k 10 --data_path '../data' --p_semi 0.2

will have error.

Traceback (most recent call last): File "main.py", line 524, in main() File "main.py", line 201, in main main_worker(args.gpu, ngpus_per_node, args) File "main.py", line 257, in main_worker train_loader, val_loader, train_sampler = get_loader(args, {'train': train_dataset, 'val': val_dataset}) File "main.py", line 448, in get_loader train_sup_dataset = train_dataset['SUP'] KeyError: 'SUP'

Is the code fit for all the dataset? and What's the reason you deal different dataset with different way? Thanks

[FriedRonaldo]

Hi, AFHQ-CAT dataset does not have the domain lables, therefore, it can not conduct the semi-supervised learning. I am so sorry for the inconvenience. It should be changed into "python main.py --gpu 0 --dataset afhq_cat --output_k 10 --data_path '../data' --p_semi 0.0"

I will change ReadMe. Thank you for letting me know that!

[ZYJ-JMF]

Thanks for your answer. I saw you used the animal faces dataset and FOOD-101 dataset. I can find the name of the class name of Animal Faces in code,but I can not find the FOOD-101 class name could you please provide the ten classes' names that used for FOOD-101 dataset?

[FriedRonaldo]

This source code is under the refactoring, so I provide the class names directly: baby_back_ribs beef_carpaccio beignets bibimbap caesar_salad clam_chowder dumplings edamame spaghetti_bolognese strawberry_shortcake

These are the file of the used classes in Food-101.

---

I copy and paste the folders and use [i for i in range(10)]. Then, give the directory of Food-10.

[ZYJ-JMF]

Thanks a lot !

[ZYJ-JMF]

Hi,could you share the set about the Food-10 dataset? Such as in your validation.py L52-61,you set this:

Parse images for average reference vector

x_each_cls = []
if args.dataset == 'animal_faces':
    num_tmp_val = -50
elif args.dataset == 'ffhq':
    num_tmp_val = -7000
elif args.dataset == 'lsun_car':
    num_tmp_val = -10000
else:
    num_tmp_val = 0

if I want to use food-10 to train,what should I set? Thanks

[FriedRonaldo]

The Food-10 dataset is basically the same as 'animal_faces' dataset. Therefore, you should change the conditions containing 'animal_faces' into a condition having 'food-10'.

For the above example, you should change to

if args.dataset in ['animal_faces', 'food-10'] 

'food-10' might be changed as you set.

[ZYJ-JMF]

ok~ got it~Thanks

[FriedRonaldo]

If it is ok, I will upload the source code without refactoring but simply applied "food-10"

[ZYJ-JMF]

oh That's cool. I will glad to see that~ Thanks again~

[FriedRonaldo]

Please check "food_added" branch. I assume that you separated a folder only for food-10.

If you have any issues, feel free to tell me.

[ZYJ-JMF]

I got it~ Thank you very much~