Add modules of SSN pipeline. #33
Conversation
Codecov Report
@@ Coverage Diff @@
## master #33 +/- ##
==========================================
+ Coverage 84.90% 85.17% +0.26%
==========================================
Files 73 73
Lines 3889 3979 +90
Branches 632 643 +11
==========================================
+ Hits 3302 3389 +87
Misses 485 485
- Partials 102 105 +3
Flags with carried forward coverage won't be shown. Click here to find out more.
Continue to review full report at Codecov.
|
| Returns: | ||
| np.ndarray: Sampled frame indices in train mode. | ||
| """ | ||
| start_frame = proposal.start_frame + 1 |
There was a problem hiding this comment.
Because the interval of the proposal is (proposal.start_frame, proposal.end_frame]
There was a problem hiding this comment.
why did they choose this convention? can we change it to the python-style interval?
JackyTown
force-pushed
the
tjq/ssn-pipeline
branch
from
d0af52f to
4d0654d
Compare
|
kindly ping |
|
Kindly ping. |
|
The interval convention part has not changed. |
JackyTown
force-pushed
the
tjq/ssn-pipeline
branch
from
0b725a6 to
b3f6166
Compare
| @@ -88,7 +89,7 @@ def _get_train_clips(self, num_frames): | |||
| ratio = (num_frames - ori_clip_len + 1.0) / self.num_clips | |||
| clip_offsets = np.around(np.arange(self.num_clips) * ratio) | |||
| else: | |||
| clip_offsets = np.zeros((self.num_clips, )) | |||
| clip_offsets = np.zeros((self.num_clips, ), dtype=np.int32) | |||
There was a problem hiding this comment.
any reason to choose int32?
There was a problem hiding this comment.
Will it be better to change it to int8？
| np.ndarray: Sampled frame indices in train mode. | ||
| """ | ||
| start_frame = proposal.start_frame | ||
| end_frame = proposal.end_frame - 1 |
There was a problem hiding this comment.
end_frame = proposal.end_frame
| end_frame = proposal.end_frame - 1 | ||
| ori_clip_len = self.clip_len * self.frame_interval | ||
|
|
||
| duration = end_frame - start_frame + 1 |
There was a problem hiding this comment.
duration = end_frame - start_frame
| assert duration != 0 | ||
| valid_length = duration - ori_clip_len | ||
|
|
||
| valid_starting = max(1, |
There was a problem hiding this comment.
is the calculation assumes that
- 1 based counting
- inclusive for both starting and ending?
i was trying to move the convention of 0 based, [start, end)
No description provided.