About Dataset Preparation issue

[lixiangyu-1008]

Hi @swathikirans thanks for your great work!

I follow the download link https://drive.google.com/file/d/1_y8Y3PnCXsngmZVMqZbg-AfJyIdOeQ2_/view?usp=sharing

But I cannot find the --trainDatasetDir ./dataset/gtea_61/split2/train in the zip file. It only have the "flow_x", "flow_y" and "frames".

[swathikirans]

I shared the frames of all the videos in the dataset. You can create the train/val splits by yourself. For example, the train data of split 2 consists of samples from S1, S3 and S4 while the val split consists of samples from S2.

[lixiangyu-1008]

Thanks for replying. So how I can set the correct the train data of split 2 as you said in your paper? Is it the train data of split 2 only sample from S2?

[swathikirans]

For split 2, copy(or make symbolic links of) dirs S1, S3 and S4 to the train dir and S2 to the test dir

[lixiangyu-1008]

Thanks, I will have a try!

[lixiangyu-1008]

Hi @swathikirans What's the different about the GTEA 61* and GTEA 61, it means that how to the fixed split?

[swathikirans]

GTEA61* is split 2 and GTEA61 in the table is the average across all four splits

[lixiangyu-1008]

Is that for GTEA61, we should setting the train/test split as follows: For split1 : training : S2,S3,S4. testing:S1 For split2 : training : S1,S3,S4. testing:S2 For split3 : training : S1,S2,S4. testing:S3 For split4 : training : S1,S2,S3. testing:S4 Then average the four split accuracy ?

[swathikirans]

Yes, you are right.

[lixiangyu-1008]

Thanks! :) Have you plan to test your model on EPIC-KITCHENS Dataset ?

[swathikirans]

No, I don't think I will try this model on EPIC-KITCHENS dataset.

[lixiangyu-1008]

Ok, Thanks, I will follow your LSTA on EPIC-KITCHENS dataset.

[frostinassiky]

@lixiangyu-1008 Any follow-up?

[chenwenhe11]

@lixiangyu-1008 I have the same problem, how did you solve it？