liubai01
commented
4 years ago BTW, I download your CUB feature, it sounds like there is a overlap of unseen test set and training set.
It may potentially be led by inconsistent labeling of matlab index(start from 1) and python index(strat from 0) from the original dataset.
>>> import scipy.io as sio
>>> data = sio.loadmat("data.mat")
>>> data.keys()
dict_keys(['__header__', '__version__', '__globals__', 'att_train', 'attribute', 'seen_pro', 'test_seen_fea', 'test_unseen_fea', 'train_fea', 'train_pro', 'unseen_pro'])
>>> label = sio.loadmat("label.mat")
>>> label.keys()
dict_keys(['__header__', '__version__', '__globals__', 'test_seen_idex', 'test_unseen_idex', 'train_idx', 'train_label_new'])
>>> np.unique(label['train_idx'])
array([ 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13,
14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26,
27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39,
40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52,
53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65,
66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78,
79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91,
92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104,
105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117,
118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130,
131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143,
144, 145, 146, 147, 148, 149, 150], dtype=uint8)
>>> np.unique(label['test_unseen_idex'])
array([ 7, 19, 21, 29, 34, 36, 50, 56, 62, 68, 69, 72, 79,
80, 87, 88, 91, 95, 98, 100, 104, 108, 116, 120, 122, 124,
125, 129, 139, 141, 142, 150, 152, 157, 159, 160, 166, 167, 171,
174, 176, 179, 182, 185, 187, 189, 191, 192, 193, 195], dtype=uint8)
>>> np.unique(label['test_seen_idex'])
array([ 1, 2, 3, 4, 5, 6, 8, 9, 10, 11, 12, 13, 14,
15, 16, 17, 18, 20, 22, 23, 24, 25, 26, 27, 28, 30,
31, 32, 33, 35, 37, 38, 39, 40, 41, 42, 43, 44, 45,
46, 47, 48, 49, 51, 52, 53, 54, 55, 57, 58, 59, 60,
61, 63, 64, 65, 66, 67, 70, 71, 73, 74, 75, 76, 77,
78, 81, 82, 83, 84, 85, 86, 89, 90, 92, 93, 94, 96,
97, 99, 101, 102, 103, 105, 106, 107, 109, 110, 111, 112, 113,
114, 115, 117, 118, 119, 121, 123, 126, 127, 128, 130, 131, 132,
133, 134, 135, 136, 137, 138, 140, 143, 144, 145, 146, 147, 148,
149, 151, 153, 154, 155, 156, 158, 161, 162, 163, 164, 165, 168,
169, 170, 172, 173, 175, 177, 178, 180, 181, 183, 184, 186, 188,
190, 194, 196, 197, 198, 199, 200], dtype=uint8)
Are the reported results in "Episode-based Prototype Generating Network for Zero-Shot Learning"