do not understand the codes in imagenet curation sub-folder

[CoinCheung]

trafficstars

Hi,

Thanks for providing this helpful codebase. I am trying to use imagenet video dataset also, and I might have to implement a python version since I use python in my project. I do not understand the details of this codebase. Will the code generate image chips with the objects in their centers, and will the generated chips be fixed in their sizes?

[bertinetto]

Hi, Yes, your interpretation is correct.

On Fri, Feb 15, 2019 at 8:56 AM CoinCheung [email protected] wrote:

Hi,

Thanks for providing this helpful codebase. I am trying to use imagenet video dataset also, and I might have to implement a python version since I use python in my project. I do not understand the details of this codebase. Will the code generate image chips with the objects in their centers, and will the generated chips be fixed in their sizes?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/bertinetto/siamese-fc/issues/69, or mute the thread https://github.com/notifications/unsubscribe-auth/ADLRuepqkb3daEfhjq4yBL3px_a6PTAnks5vNnZPgaJpZM4a9Neq .

[CoinCheung]

Thanks! In one case, the target size is 255x255, and the bounding box size of the object is 40x60, while the image size is 500x700. I can crop a chip of size 255x255 with the 40x60 bounding box at its center. However, suppose the bounding box has size 400x600, shall I still crop a 255x255 chip within the 400x600 bounding box, or do I need to resize? If so, would you please tell me what size should I resize the image into?