search

GigaVision / PANDA-Toolkit

PANDA dataset tool kit for data visualization, splitting, merging, and result evaluation.
MIT License
93 stars 24 forks source link

I have a question, please help me? #2

Closed muzishen closed 4 years ago

muzishen commented 4 years ago

Hello, I am a rookie at DOTA. I understand that the big picture is changed into many patch pictures to detecte via sliding window, and then the results are combined, right? There is a place I don’t understand, you can give an example. https://github.com/GigaVision/PANDA-Toolkit/blob/7bc91154aff724d557eeab5a0569e8d213053ada/ImgSplit.py#L71, param imgrequest and param imgfilters in test, l should how to do?

DarkstartsUp commented 4 years ago

param imgrequest is a list of images names you want to request, eg. ['1-HIT_canteen/IMG_1_4.jpg', ...] param imgfilters is essential keywords in image name, eg. ['HIT', 'OCT', ...] You may follow the usage of this function is in demo.py: https://github.com/GigaVision/PANDA-Toolkit/blob/71231cc058bc04f2ca053aca37032ada9c274990/demo.py#L34

muzishen commented 4 years ago

Thank you!

muzishen commented 4 years ago

Does the image split script of task 2 (MOT) also use this? It seems that there is not key of 'person' in the seqinfo.json file.

DarkstartsUp commented 4 years ago

For now ,image split script is only for task 1 (Object Detection on PANDA-IMAGE). However, a few changes to the data interface can make the code run on PANDA-VIDEO. We will add this part of function in the future, please modify it according to your needs at present.