hvkwak / pie-net_dataset_preparation Goto Github PK

My recently downloaded ABC Dataset is formed in the following manner:

• 100 chunks in total: 0000-0099
• each chunk contains 10000 models: 00000000-00009999, model folders are under chunk folder
• each model folder may or may not contain annotation files

Thus, for .yml file of chunk 0000, it's like:

0000_feat/00009982/00009982_8fd881580c344a9583659586_features_016.yml
0000_feat/00009986/00009986_ba508426162c40128d0c7a2a_features_000.yml
0000_feat/00009987/00009987_ba508426162c40128d0c7a2a_features_001.yml
0000_feat/00009988/00009988_ba508426162c40128d0c7a2a_features_002.yml
0000_feat/00009989/00009989_ba508426162c40128d0c7a2a_features_003.yml
0000_feat/00009993/00009993_0dc78076fcce4a67864e71d4_features_000.yml
0000_feat/00009994/00009994_ea83e2f413614058bcfa17e1_features_000.yml
0000_feat/00009996/00009996_6980d41656c4406383dc8b46_features_001.yml
0000_feat/00009998/00009998_6980d41656c4406383dc8b46_features_003.yml
0000_feat/00009999/00009999_003df34c1a7b4dc2bd43c5f4_features_000.yml

In line 19 of run_make_list.sh: cp $dir_feat/${line1}/*.yml ${line1}

It seems to me that run_make_list.sh cannot handle this correctly. The code seems to assume that all annotation files are located directly under the chunk folder, e.g. 0000_feat/xxx_features_xxx.yml.

So my question: Was the ABC Dataset updated since the creation of this code, or did I misunderstood something? If the dataset organization is changed, what else do I need to change besides the run_make_list.sh?

Hello, when I run main2.py, it suggests that the file.json file is missing, I hope you can provide the file.
And, according to my analysis, the code is used to synthesise the data of multiple MAT files to batch_count=64, which means that each MAT file contains 64×1 struct in Training_data, but the actual result does not seem to be the case.