Skip to content
This repository has been archived by the owner on Nov 17, 2023. It is now read-only.

Onnx Fix 6 MaskRCNN models #20178

Merged
merged 4 commits into from
Apr 19, 2021
Merged

Onnx Fix 6 MaskRCNN models #20178

merged 4 commits into from
Apr 19, 2021

Conversation

Zha0q1
Copy link
Contributor

@Zha0q1 Zha0q1 commented Apr 15, 2021

Fixes 6 maskrcnn models

mask_rcnn_resnet18_v1b_coco
mask_rcnn_fpn_resnet18_v1b_coco
mask_rcnn_resnet50_v1b_coco
mask_rcnn_fpn_resnet50_v1b_coco
mask_rcnn_resnet101_v1d_coco
mask_rcnn_fpn_resnet101_v1d_coco

TODO: add model tests
mx2onnx_maskrcnn

@Zha0q1 Zha0q1 requested a review from szha as a code owner April 15, 2021 23:21
@mxnet-bot
Copy link

Hey @Zha0q1 , Thanks for submitting the PR
All tests are already queued to run once. If tests fail, you can trigger one or more tests again with the following commands:

  • To trigger all jobs: @mxnet-bot run ci [all]
  • To trigger specific jobs: @mxnet-bot run ci [job1, job2]

CI supported jobs: [windows-cpu, unix-cpu, sanity, windows-gpu, centos-gpu, miscellaneous, unix-gpu, clang, website, centos-cpu, edge]


Note:
Only following 3 categories can trigger CI :PR Author, MXNet Committer, Jenkins Admin.
All CI tests must pass before the PR can be merged.

@Zha0q1 Zha0q1 changed the title [wip] Onnx Fix 6 MaskRCNN models Onnx Fix 6 MaskRCNN models Apr 16, 2021
@@ -240,14 +238,20 @@ def obj_detection_test_images(tmpdir_factory):
'faster_rcnn_resnet101_v1d_coco',
'yolo3_darknet53_coco',
'yolo3_mobilenet1.0_coco',
'mask_rcnn_resnet18_v1b_coco',
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

By adding the mask_rcnn under obj_detection tests, we are not testing the segmentation part?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think the mask is a little hard to test numerically since they can be different across mxnet and onnxruntime

@@ -199,8 +199,6 @@ def obj_detection_test_images(tmpdir_factory):
tmpdir = tmpdir_factory.mktemp("obj_det_data")
from urllib.parse import urlparse
test_image_urls = [
'https://github.com/apache/incubator-mxnet-ci/raw/master/test-data/images/car.jpg',
'https://github.com/apache/incubator-mxnet-ci/raw/master/test-data/images/duck.jpg',
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why do we remove the images here?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Those two images would fail the maskrcnn models. It's not like maskrcnn models do not work on them, but they would output slightly different boxes than mxnet

Copy link
Contributor

@waytrue17 waytrue17 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM, thanks!

@Zha0q1 Zha0q1 merged commit 4bd7ad5 into apache:v1.x Apr 19, 2021
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants