coco converter #274
coco converter #274
Conversation
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
| segmentation.extend([ | ||
| box.start.x, box.start.y, box.end.x, box.start.y, box.end.x, | ||
| box.end.y, box.start.x, box.end.y | ||
| ]) |
There was a problem hiding this comment.
q: does the orientation matter for coco?
| def rle_to_common(class_annotations, class_name): | ||
| mask = rle_decoding(class_annotations.segmentation.counts, | ||
| *class_annotations.segmentation.size[::-1]) | ||
| return ObjectAnnotation(name=class_name, |
There was a problem hiding this comment.
Not I was going to leave a comment "I don't see name in ObjectAnnotation but then realized that BaseAnnotation inherits from FeatureSchema. Imho it's quite confusing, it should be a composition rather than inheritance.
| futures = [] | ||
| coco_categories = {} | ||
|
|
||
| with ProcessPoolExecutor(max_workers=8) as exc: |
There was a problem hiding this comment.
let's make executor an optional argument of from_common
There was a problem hiding this comment.
See question below for same comment on panoptic dataset.
| coco_categories = {} | ||
| coco_things = {} | ||
| images = [] | ||
| with ProcessPoolExecutor(max_workers=8) as exc: |
There was a problem hiding this comment.
Same comment about executor as arg. Not being able to specify concurrency would be pretty annoying to me as user
There was a problem hiding this comment.
What if I just let the number of workers be an arg? My thinking is that most users are going to want the same amount of parallelism - the max that the backend can handle. Also this script is unlikely to run within the context of another executor - users are going to either convert an entire dataset or not. So asking users to pass in the execturor is an extra step that will always be necessary.
|
I changed everything to use Pathlib - including the public interfaces. I am a little bit hesitant to keep it that way because we don't support that in other interfaces. Let's chat about this. |
…labelbox-python into ms/custom-metric-tutorial
…labelbox-python into ms/custom-metric-tutorial
…labelbox-python into ms/custom-metric-tutorial
No description provided.