What if I'd like to use own dataset

[mhyeonsoo]

Thanks for great sources.
I want to try to implement this with my own dataset, so I'd like to ask you guys which settings or configuration I should change.

What I am thinking of are, config file and dataset path (annotated in Pascal VOC ideally).
Is there any other things that I should take care of?

Thank you.

[AruniRC]

Hi, there are a couple of steps needed to do this: 1) First, please go thru the "dataset catalog", which lists all the available datasets, annotations and paths https://github.com/AruniRC/detectron-self-train/blob/master/lib/datasets/dataset_catalog.py (a new dataset needs to be added to this catalog) Main things to note are, each dataset consists of an IM_DIR, a root folder where images are saved, and a ANN_FN or annotation file, in JSON format, which contains training annotations such as bounding boxes, class labels and paths to images (these image paths are relative to IM_DIR). 2) Creating the JSON annotations file (ANN_FN needs to point to this). The Detectron training pipeline accepts annotations in a particular JSON format compatible with MS-COCO. It's best to look at a few examples to get an idea on how to do this: (i) converting face detection annotations (bboxes) to json: https://github.com/AruniRC/detectron-self-train/blob/master/lib/datasets/wider/convert_face_to_coco.py (ii) Converting Cityscapes dataset into COCO format JSON: https://github.com/AruniRC/detectron-self-train/blob/master/lib/datasets/cityscapes/cityscapes_to_json.py hope that helps~! Aruni Roy Chowdhury http://people.cs.umass.edu/~arunirc/

…

On Thu, Jan 2, 2020 at 9:15 PM Hyeonsoo Moon ***@***.***> wrote: Thanks for great sources. I want to try to implement this with my own dataset, so I'd like to ask you guys which settings or configuration I should change. What I am thinking of are, config file and dataset path (annotated in Pascal VOC ideally). Is there any other things that I should take care of? Thank you. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#14?email_source=notifications&email_token=AAIBQT5SD3UEORDI3KPOE5DQ32NSPA5CNFSM4KCIX5A2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4IDZLN4A>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAIBQT4TNZQRB2EPBH2FYZTQ32NSPANCNFSM4KCIX5AQ> .

[mhyeonsoo]

Dataset catalog guides me perfectly.
I will try following this script.
Thanks a lot for quick response!

[GiovanniPasq]

Which file yaml i need to change to select my datasets ? which config i need to run with python?
thanks