VOC dataset question

[rudylyh]

Hello! Thanks for sharing your code. I noticed that your setting on the VOC dataset is different from the PANet (https://github.com/kaixin96/PANet). Their training data and test data are from the training set. Your training data are 15 classes from the training set and test data are 5 classes from the validation set. The numbers of images in each class are also different. May I ask why there are two different settings? Are the performances on two settings comparable? Thank you in advance.

[tianzhuotao]

@rudylyh Thanks for your interest in our work!

We follow the descriptions in OSLSM to build our dataset where the images for validation should be collected from the validation set of Pascal VOC. You can refer to the paper of OSLSM for more details. In OSLSM, Shaban et al wrote that (at the end of section 5):

"We follow a similar procedure to
form the test set D_test, but here the image-label pairs are taken from PASCAL VOC validation set and the corresponding label-set is L_test".

Also, we follow the instructions available on SBD:

"Please note that the train and val splits included with this dataset are different from the splits in the PASCAL VOC dataset. In particular some "train" images might be part of VOC2012 val".

"If you are interested in testing on VOC 2012 val, then use this train set, which excludes all val images. This was the set used in our ECCV 2014 paper. This train set contains 5623 images."

Therefore we use the collection that excludes all images contained in the validation set as our training set, and we have shown the procedure to yield the data lists in this repo.

We have tried to reproduce PANet on our dataset and the results are comparable. Of course, you can accordingly modify the code to have fair comparisons.

Hope my answer addresses your questions : )

[rudylyh]

Thanks for your detailed answer. It helps me a lot.