dataset in table 5

[freedom6927]

Hi, could you release the dataset in Table5, thank you.

[TheShadow29]

@freedom6927 please see https://github.com/TheShadow29/zsgnet-pytorch/blob/master/DATA_README.md

You need to download the annotation files in the drive link.

Let me know if you run into any error.

[freedom6927]

Hi, I have downloaded. But I can't distinguish which files are related to table 5. I found all the files are related to table 2-4 Arka Sadhu ***@***.***> 于2022年2月11日周五 02:40写道：

…

@freedom6927 <https://github.com/freedom6927> please see https://github.com/TheShadow29/zsgnet-pytorch/blob/master/DATA_README.md You need to download the annotation files in the drive link. Let me know if you run into any error. — Reply to this email directly, view it on GitHub <#13 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AKGZSHTHIZEN6GMZT25T3ELU2QBBRANCNFSM5N7UJO3A> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[TheShadow29]

@freedom6927 it is under vg_split / csv_dir.

For training and validation, you could use training_balanced.csv and val_balanced.csv

For testing, you would use test_balanced_c2.csv and test_balanced_c3.csv respectively.

In table 5, we train on train_balanced, and validate using val_balanced and finally test it on test_balanced_c2.csv(VG-2B) andtest_balanced_c3.csv` (VG-3B)

For the distances, you would need the object name from test_...csv file and find the closest object in the training set and then simply bucket them into 3-4, 4-5 and so on, and get results for those each subset.

Let me know if that answers your question.

[freedom6927]

Got it, thanks! Arka Sadhu ***@***.***> 于2022年2月12日周六 07:13写道：

…

@freedom6927 <https://github.com/freedom6927> it is under vg_split / csv_dir. For training and validation, you could use training_balanced.csv and val_balanced.csv For testing, you would use test_balanced_c2.csv and test_balanced_c3.csv respectively. In table 5, we train on train_balanced, and validate using val_balanced and finally test it on test_balanced_c2.csv(VG-2B) andtest_balanced_c3.csv` (VG-3B) For the distances, you would need the object name from test_...csv file and find the closest object in the training set and then simply bucket them into 3-4, 4-5 and so on, and get results for those each subset. Let me know if that answers your question. — Reply to this email directly, view it on GitHub <#13 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AKGZSHSRX7I3TW2RAZNZ4ULU2WJ2ZANCNFSM5N7UJO3A> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[freedom6927]

Hi, I have a new question in table 4. are the datasets of unblance test_c2.csv, test_c3.csv and train.csv. But where is the val.csv. The only found of val is val_balanced.csv Arka Sadhu ***@***.***> 于2022年2月12日周六 07:13写道：

…

@freedom6927 <https://github.com/freedom6927> it is under vg_split / csv_dir. For training and validation, you could use training_balanced.csv and val_balanced.csv For testing, you would use test_balanced_c2.csv and test_balanced_c3.csv respectively. In table 5, we train on train_balanced, and validate using val_balanced and finally test it on test_balanced_c2.csv(VG-2B) andtest_balanced_c3.csv` (VG-3B) For the distances, you would need the object name from test_...csv file and find the closest object in the training set and then simply bucket them into 3-4, 4-5 and so on, and get results for those each subset. Let me know if that answers your question. — Reply to this email directly, view it on GitHub <#13 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AKGZSHSRX7I3TW2RAZNZ4ULU2WJ2ZANCNFSM5N7UJO3A> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[TheShadow29]

@freedom6927 sorry for the late reply, we only used the balanced validation set.

[freedom6927]

Ok, final question. Could you release the dataset division of Table 5 to me? I hold refer to your designed ideas to finish my research. Arka Sadhu ***@***.***> 于2022年2月15日周二 06:12写道：

…

@freedom6927 <https://github.com/freedom6927> sorry for the late reply, we only used the balanced validation set. — Reply to this email directly, view it on GitHub <#13 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AKGZSHS2MEVWKUX6JHW4WWTU3F44NANCNFSM5N7UJO3A> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[TheShadow29]

@freedom6927 Sorry, I don't understand your question. What do you mean by dataset division?

[freedom6927]

[image: table5.png] you see. In table 5. The testset of vg2B and vg3B are divided into 5 parts by using GloVe embeddings to compute the distances between test sample and training sample. Could you release the results of division to me (the details of 310, 1050, 3543, 5321, 1985 of vg2B) Arka Sadhu ***@***.***> 于2022年2月15日周二 11:52写道：

…

@freedom6927 <https://github.com/freedom6927> Sorry, I don't understand your question. What do you mean by dataset division? — Reply to this email directly, view it on GitHub <#13 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AKGZSHQQCV2MZO3ZOOQOEXDU3HEZDANCNFSM5N7UJO3A> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[TheShadow29]

@freedom6927 I don't have it with me, but something along the following lines of code should be sufficient:

train_df = ... # read train csv
test_df = ... # read test csv

train_objects = train_df['object_name'].unique()
test_objects = train_df['object_name'].unique()

glove_emb = ... # read glove embeddings

train_obj_emb = glove_emb(train_objects)
test_obj_emb = glove_emb(test_objects)

test_dist_dict = {}
for test_obj in test_objects:
    # find closest train object
    closest_train_obj = ....
    closest_train_obj_dist = ....
    test_dist_dict[test_obj] = closest_train_obj_dist

# bucket by distances

# compute scores for each object

Let me know if this answers your question.

[freedom6927]

thanks very much. I understand your idea Arka Sadhu ***@***.***> 于2022年2月15日周二 12:35写道：

…

@freedom6927 <https://github.com/freedom6927> I don't have it with me, but something along the following lines of code should be sufficient: train_df = ... # read train csv test_df = ... # read test csv train_objects = train_df['object_name'].unique() test_objects = train_df['object_name'].unique() glove_emb = ... # read glove embeddings train_obj_emb = glove_emb(train_objects) test_obj_emb = glove_emb(test_objects) test_dist_dict = {} for test_obj in test_objects: # find closest train object closest_train_obj = .... closest_train_obj_dist = .... test_dist_dict[test_obj] = closest_train_obj_dist # bucket by distances # compute scores for each object Let me know if this answers your question. — Reply to this email directly, view it on GitHub <#13 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AKGZSHR3QLKE2INZPEHOVILU3HJZDANCNFSM5N7UJO3A> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>. You are receiving this because you were mentioned.Message ID: ***@***.***>