ICDAR 2013 - Gender Identification Competition Dataset (GenderIdentifify2013)
Ground Truth
Genders of all writers
Keywords
Gender recognition
Description
The genders of the writers of the training set are in train_answers.csv and those of the writers of the test set are in the test_answers.csv
Important Note: This competition has been hosted in kaggle. This website allows participants to see how well
each of their submissions performs on a part of the test set called the public test set (37% for this competition).
However, only the remaining part of the test set (called the private test set) counts towards the final standing. Therefore,
the ranking on the public test set is not necessarily similar to that of the private test set. This is mainly due to participants overfitting to the public test set. In the test_answers.csv there is a column called Usage, this column indicates for each row whether it is used in the public or private leaderboard. Consequently, in order to have results which are comparable with those of the competition, the public test set must be used as a validation set and the private test set as a final test set.
| File | Type | Size | Downloads | Description |
|---|---|---|---|---|
| train_answers.csv | data | (2 KB) | 35 | Genders of writers of the training set |
| test_answers.csv | data | (4 KB) | 33 | Genders of writers of the test set |
Comments
Add your comment
In order to comment on a dataset you need to be logged on
Valoration
Recently I have started to engage with this field and built a CNN model in order to predict gender through offline handwriting.
I would like to evaluate my performance on the QUWI subset, used in the competition. There is any source to get an output example with the male scores? I have my model's scores output buy after calculation the LogLoss score my results seems to be very good. I want an example of full output in order to compare the calculation process.
Thanks in advance,
Evya.