search

leokarlin / LaSO

LaSO: Label-Set Operations networks for multi-label few-shot learning - official implementation
BSD 3-Clause "New" or "Revised" License
86 stars 14 forks source link

Does anyone run "test_ augmentation.py ”successfully? #12

Closed mohong96 closed 4 years ago

mohong96 commented 4 years ago

1.when i use the command line：“python test_augmentation.py --base_network_name='Inception3' --batch_size=4 --class_cap=5 --class_ind_dict_path='/data_for_augmentation/5shotRun1ClassIdxDict.pkl' --classifier_name='Inception3Classifier' --crop_size=299 --g_inner_dim=2048 --init_inception=1 --latent_dim=2048 --lr=0.01 --n_epochs=50 --paper_reproduce=1 --resume_path= --sets_basic_block_name='SetopResBasicBlock' --sets_block_name='SetopResBlock_v1' --sets_network_name='SetOpsResModule' --used_ind_path='/data_for_augmentation/5shotRun1UsedIndices.pkl' --results_path= --coco_path=” The error is：test_augmentation.py: error: unrecognized arguments: --base_network_name=Inception3 --classifier_name=Inception3Classifier --g_inner_dim=2048 --init_inception=1 --paper_reproduce=1 --sets_basic_block_name=SetopResBasicBlock --sets_block_name=SetopResBlock_v1 --sets_network_name=SetOpsResModule image

mohong96 commented 4 years ago
  1. i do not know how to find "/dccstor/alfassy/saved_models/FVClassAug5AVG2018.10.30.15:32usedIndicesRun4.pkl' 、‘'/dccstor/alfassy/saved_models/FVClassAug5AVG2018.10.30.15:32ClassIdxDict16Run4.pkl' and “/dccstor/alfassy/finalLaSO/code_release/trainedModels” image image
mohong96 commented 4 years ago

I'm very interested in laso, but I don't know how to solve the above two problems. If anyone has a solution, please let me know. Thank you

zscd commented 4 years ago

I am confused about this too

alfassy commented 4 years ago

Hi,

I fixed the README of this repository to have a working run command for augmentation, thank you for pointing our the bug. for "used_ind_path" and "class_ind_dict_path" you need to use the files from the "data_for_augmentation" folder, look at the example run command. you are supposed to use your own local paths, for me it was /dccstor/alfassy/... . The models can be downloaded through the google drive link in the READEME