Data format for training

[jonbakerfish]

During make runtest, set_id_to_class_file("/afs/cs.stanford.edu/u/vigneshr/scratch/ICCV2015/data/_temp/dummy.txt") is required. Where can I download this file or how does it look like?

The med_full_devtest_sampled_frames_min5_max1000_lmdb, med_test_event_kit_shot_c4_for_test_sample0 and test_event_kit_id_to_class.txt are required for training. Where can I download these file or how do they look like?

[eevignesh]

I will fix the path to the file. Meanwhile, you can download the file from: http://ai.stanford.edu/~vigneshr/misc/dummy.txt

[jonbakerfish]

@eevignesh Thanks for the link.

Can you explain a little bit how do the lmdb files for training look like? Or how can I construct these files on my own dataset. Is there any code in this repo can be reused?

[romilbhardwaj]

@eevignesh As requested by @jonbakerfish, it would be very helpful if you could explain a bit about how the lmdb training files are structured and if there are any scripts we can reuse to convert our video data into the format required by VIDEO_SAMPLED_SHOTS_DATA layer.

@jonbakerfish, did you manage to figure out something about the format?

[ailanchong]

@romilbhardwaj did you manage to figure out something about the format?

[ailanchong]

No，I gave up. | | 陈鹏达 | | 邮箱：[email protected] | 签名由 网易邮箱大师 定制 On 03/25/2019 20:49, wangxiaohuabuaa wrote: @ailanchong did you manage to figure out something about the format? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.

[Flowerhwang]

@romilbhardwaj did you manage to figure out something about the format?