Mask_rcnn: AssertionError

Created on 20 Jul 2018  路  2Comments  路  Source: matterport/Mask_RCNN

I am running inspect_balloon_model on my machine and got assertion error at 1.b RPN Predictions:
For me it looks close to issue #23 , but solution from it doesn't help. nms_node is empty.
There could be difference in tf versions. My setup: tf 1.9, keras 2.1.6, cuda 9.2, nvidia driver 396.37, Ubuntu 16.04. The only one change in the model by me was adding that 'print' on the screenshot.

default

Hope, you could give me an advice.
Thanks in advance

Most helpful comment

In case any else comes accross this: In tensorflow 1.9 it is now called NonMaxSuppressionV3:0, so the code should be:

nms_node = model.ancestor(pillar, "ROI/rpn_non_max_suppression:0")
if nms_node is None:
    nms_node = model.ancestor(pillar, "ROI/rpn_non_max_suppression/NonMaxSuppressionV2:0")
if nms_node is None:
    nms_node = model.ancestor(pillar, "ROI/rpn_non_max_suppression/NonMaxSuppressionV3:0")

All 2 comments

The issue was solved by rollback of tf to 1.8 and cuda to 9.0.

In case any else comes accross this: In tensorflow 1.9 it is now called NonMaxSuppressionV3:0, so the code should be:

nms_node = model.ancestor(pillar, "ROI/rpn_non_max_suppression:0")
if nms_node is None:
    nms_node = model.ancestor(pillar, "ROI/rpn_non_max_suppression/NonMaxSuppressionV2:0")
if nms_node is None:
    nms_node = model.ancestor(pillar, "ROI/rpn_non_max_suppression/NonMaxSuppressionV3:0")
Was this page helpful?
0 / 5 - 0 ratings