search

Cheng-Lin-Li / SegCaps

A Clone version from Original SegCaps source code with enhancements on MS COCO dataset.
Apache License 2.0
65 stars 29 forks source link

Creating MSCOCO dataset #10

Closed shan18 closed 5 years ago

shan18 commented 5 years ago

How did you create the label vector (y) when using the model on MSCOCO dataset? For each input (image), MSCOCO has multiple outputs (categories). How to create the dataset in this case (i.e. the input vector x and the label vector y)? Any help would be appreciated. Thank You.

Cheng-Lin-Li commented 5 years ago

Hi Shan18,

You may try to look into my implementation of MSCOCO crawler. https://github.com/Cheng-Lin-Li/SegCaps/blob/master/cococrawler/getcoco17.py

pycocotools.coco provide API to let you identify the annotation for each category of an instance. My implementation performs binary classification, the output mask file (y) only contain 0 or 1 for each pixel. You can define a value index to match with your categories.

shan18 commented 5 years ago

Okay, thank you for the help.