Commit 8904ea8a authored by uoega's avatar uoega
Browse files

clean up

parent 05d655ba
Sample 9
NEW TRAININGS SESSION with NTU_RelationNet_multi_label
Properties:
Dataset: stgcn_40_class_epoch80_features_reindex_norm
Language embedding: ['sentence_40_mean_ver2_norm', 'sentence_40_mean_ver4_norm', 'sentence_40_mean_ver5_norm']
Unseen Classes: [5, 17, 19, 24, 27, 45, 57, 59, 64, 67, 85, 97, 99, 104, 107]
Seed: 23
Batchsize: 128
Dropout: 0.5
Bias: True
Batchnorm: False
Weight init with xavier: False
Multi hot: False
Inferenz with sum of relation scores: False
Dimension vom Language Embedding: (30720,)
Dimension vom Language Embedding: (61440,)
Dimension vom Language Embedding: (92160,)
Dimension vom Language Embedding: (120, 768)
Unseen Classes: [5, 17, 19, 24, 27, 45, 57, 59, 64, 67, 85, 97, 99, 104, 107]
Using Model: 500k
GZSL Unseen Class Accuracy:
Label: [[27]]
Top 5: [ 27 2 42 82 107]
[0.4149502 0.35621354 0.26102823 0.25498113 0.1207521]
Top 15 Modulo unique: [27. 2. 36. 35. 12. 39.]
Adapted relation scores: [0.91782318 0.9121652 0.79746223 0.71359468 0.66594639 0.66278358]
Relation scores: [0.41495019 0.35621354 0.02195648 0.00315389 0.00105174 0.00097783]
top 5 acc for class: 27 1.0
Accuracy for class 27 : 1.0
Confusion Scores for Class 27 (>10%):
GZSL Seen Class Accuracy:
Label: [[22]]
Top 5: [ 22 102 62 24 25]
Top 15 Modulo unique: [22. 24. 25. 19. 20. 23. 10.]
Adapted relation scores: [0.89612739 0.88690615 0.88533399 0.82322471 0.81586078 0.81375734
0.81168559]
Relation scores: [0.23443718 0.18568955 0.17853129 0.04008907 0.03373242 0.03211224
0.03059368]
top 5 acc for class: 22 1.0
Accuracy for class 22 : 1.0
Confusion Scores for Class 22 (>10%):
gzsl: seen=1.0000, unseen=1.0000, h=1.0000
gzsl top5: seen=1.0000, unseen top5=1.0000, h=1.0000
NEW TRAININGS SESSION with NTU_zsar40
Properties:
Dataset: stgcn_40_class_epoch80_features_reindex_norm
Language embedding: language_embedding_40_bert_mean_norm
Unseen Classes: [5, 17, 19, 24, 27]
Seed: 23
Dropout: 0.5
(env_st-gcn) C:\Users\XPS15\Documents\Eigene Dokumente\Uni\Master\zero-shot-action-recognition\LearningToCompare_ZSL>"c:/Users/XPS15/Documents/Eigene Dokumente/Uni/Master/zero-shot-action-recognition/st-gcn/env_st-gcn/Scripts/python.exe" "c:/Users/XPS15/Documents/Eigene Dokumente/Uni/Master/zero-shot-action-recognition/LearningToCompare_ZSL/NTU_RelationNet_test_single_sample.py"
Unseen Classes: [5, 17, 19, 24, 27]
Language embedding: language_embedding_40_bert_mean_norm
Using Model: 500k
GZSL Unseen Class Accuracy:
Label: [[27]]
[[ 2 27 36 35 17]]
Adapted relation scores: [0.94392101 0.81732137 0.68269 0.66622576 0.6623002 ]
Relation scores: [0.88331246 0.03490609 0.00154686 0.00105853 0.000967 ]
top 5 acc for class: 27 1.0
Accuracy for class 27 : 0.0
Confusion Scores for Class 27 (>10%):
Confusion with class: 2 (proportion): 1.0
GZSL Seen Class Accuracy:
Label: [[22]]
[[22 24 25 5 19]]
Adapted relation scores: [0.94284933 0.92247968 0.90267141 0.86377904 0.80653934]
Relation scores: [0.85447633 0.47164091 0.27745107 0.10517248 0.02712905]
top 5 acc for class: 22 1.0
Accuracy for class 22 : 1.0
Confusion Scores for Class 22 (>10%):
gzsl: seen=1.0000, unseen=0.0000, h=0.0000
gzsl top5: seen=1.0000, unseen top5=1.0000, h=1.0000
__________________
Descriptive
__________________
nseen Classes: [5, 17, 19, 24, 27]
Using Model: 500k
GZSL Unseen Class Accuracy:
Label: [[27]]
[[ 2 27 35 12 13]]
Adapted relation scores: [0.94597788 0.92846303 0.85434461 0.7689299 0.72443867]
Relation scores: [0.9420315 0.55807549 0.08380191 0.01132301 0.00405024]
top 5 acc for class: 27 1.0
Accuracy for class 27 : 0.0
Confusion Scores for Class 27 (>10%):
Confusion with class: 2 (proportion): 1.0
\ No newline at end of file
......@@ -15,11 +15,11 @@ import torch.nn.functional as F
parser = argparse.ArgumentParser(description="Zero Shot Learning")
parser.add_argument("-g","--gpu",type=int, default=0)
parser.add_argument("-u","--unseen_classes", type = int, nargs = '*', default = [32, 9, 34])
parser.add_argument("-s","--semantic_input",type = str, nargs = '*', default = ['sentence_40_mean_ver2_norm', 'sentence_40_mean_ver3_norm', 'sentence_40_mpnet_ver4n1_norm' ])
parser.add_argument("-d","--dropout",type=float, default=0.25)
parser.add_argument("-s","--semantic_input",type = str, nargs = '*', default = ['sentence_40_mean_ver2_norm', 'class_sentences_40_ver2_roberta_aug_1_norm', 'class_sentences_40_ver2_roberta_aug_2_norm', 'class_sentences_40_ver2_roberta_aug_3_norm', 'class_sentences_40_ver2_roberta_aug_4_norm', 'sentence_40_mean_ver5_norm', 'class_sentences_40_ver5_roberta_aug_1_norm', 'class_sentences_40_ver5_roberta_aug_2_norm', 'class_sentences_40_ver5_roberta_aug_3_norm', 'class_sentences_40_ver5_roberta_aug_4_norm', 'sentence_40_mean_ver4_norm', 'class_sentences_40_ver4_roberta_aug_1_norm', 'class_sentences_40_ver4_roberta_aug_2_norm', 'class_sentences_40_ver4_roberta_aug_3_norm', 'class_sentences_40_ver4_roberta_aug_4_norm'])
parser.add_argument("-d","--dropout",type=float, default=0)
parser.add_argument("--log_filename",type=str, default="log.txt")
parser.add_argument("--input_dir",type=str, default='session_2021-07-05_13_55_42')
parser.add_argument("--input_model",type=str, default='best_zsl')
parser.add_argument("--input_dir",type=str, default='session_2021-07-14_23_08_50')
parser.add_argument("--input_model",type=str, default='500k')
args = parser.parse_args()
......@@ -130,8 +130,8 @@ def main():
elif INPUT_MODEL == "500k":
print("Using Model: ", INPUT_MODEL)
attribute_network.load_state_dict(torch.load(INPUT_DIR + "/NTU_zsar40_attribute_network_500000.pkl"))
attribute_network.load_state_dict(torch.load(INPUT_DIR + "/NTU_zsar55_5_attribute_network_500000.pkl", map_location=torch.device('cpu')))
else:
print("Input Model ----",INPUT_MODEL,"---- not existing!")
exit()
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment