What is meaning of 'evalutation-flags' ?

[davsolai]

Thank you for your work.

When I run 'test_net.py', GeneralizedRCNN in generalized_rcnn.py return list with len=4.

and each 'list' (4 number of) in returned list has different num_boxes.

When I revise augument '--evalutation-flags' from [1, 1, 1, 1] (default) to [1], returned result has 4 lists.

What is meaning of 'evalutation-flags' ? I just want to use 'double-head-ext' only.

[AliceAsky]

I have the same needs, and I want to know what are the four differences.

[wuyuebupt]

@SoYeol-Lee Sorry that I thought I have replied until I received the next comment.

Where to use the evaluation flag can be found at the code:
https://github.com/wuyuebupt/doubleheadsrcnn/blob/master/maskrcnn_benchmark/modeling/roi_heads/box_head/inference.py#L190

The mapping from the four flags to the configuration are:
#### return multiple post processor with different mode
# 0 : conv cls + conv reg
# 1 : fc cls + fc cls
# 2 : fc cls + conv reg
# 3 : fc cls + conv reg (double-head-ext)
#------
# evaluation_flags: 1 1 1 1

So if you only need to run the 'double-head-ext'. You should use the flag "0 0 0 1".

The detailed process can be found from

doubleheadsrcnn/maskrcnn_benchmark/modeling/roi_heads/box_head/inference.py

Line 71 in a744b41

if self.mode == 0:

to

doubleheadsrcnn/maskrcnn_benchmark/modeling/roi_heads/box_head/inference.py

Line 85 in a744b41

box_regression = box_regression_conv

[wuyuebupt]

@AliceAsky check my comment above.