Reproducing results in the paper about semi-supervised-learning

[qianlanwyd]

Yes, we have modified the codes and hyper-parameters(For example, we will use layer-wise learning rate for better results in the new version and upload codes for SOTA imbalanced SSL algorithms).
We will upload new codes and logs/model checkpoints very soon. Besides, we will modify the results in the paper and will continuously update results if codes modified.

[vinnamkim]

@qianlanwyd

Hi, thanks for the reply, I hope it will be soon. Besides, can I ask about a text file ( './data/semi_fgvc/semi_aves/l_train_val.txt') required for Semi-AVES dataset? I downloaded the dataset from https://www.kaggle.com/c/semi-inat-2020, but I could not find this file anywhere.

Traceback (most recent call last):
  File "train.py", line 364, in <module>
    main(args)
  File "train.py", line 73, in main
    main_worker(args.gpu, ngpus_per_node, args)
  File "train.py", line 118, in main_worker
    dataset_dict = get_dataset(args, args.algorithm, args.dataset, args.num_labels, args.num_classes, args.seed, args.data_dir)
  File "/local_ssd3/vinnamki/Semi-supervised-learning/semilearn/utils.py", line 117, in get_dataset
    lb_dset, ulb_dset, eval_dset = get_semi_aves(args, algorithm, dataset, train_split='l_train_val', data_dir=data_dir)
  File "/local_ssd3/vinnamki/Semi-supervised-learning/semilearn/datasets/cv_datasets/aves.py", line 47, in get_semi_aves
    train_labeled_dataset = iNatDataset(alg, data_dir, train_split, dataset, transform=transform_weak)
  File "/local_ssd3/vinnamki/Semi-supervised-learning/semilearn/datasets/cv_datasets/aves.py", line 103, in __init__
    self.samples, self.num_classes, self.targets = make_dataset(self.dataset_root,
  File "/local_ssd3/vinnamki/Semi-supervised-learning/semilearn/datasets/cv_datasets/aves.py", line 62, in make_dataset
    with open(split_file_path, 'r') as f:
FileNotFoundError: [Errno 2] No such file or directory: './data/semi_fgvc/semi_aves/l_train_val.txt'

[YUE-FAN]

Hi vinnamkim,

We followed "A Realistic Evaluation of Semi-Supervised Learning for Fine-Grained Classification" for the dataset construction. The required file can be found here.

[Hhhhhhao]

Updated in fb36c1c