Simplify video_domain_adapter #292
Conversation
xianyuanliu
added
enhancement
Codecov Report
@@ Coverage Diff @@
## main #292 +/- ##
==========================================
- Coverage 92.98% 91.10% -1.89%
==========================================
Files 45 45
Lines 4635 4790 +155
==========================================
+ Hits 4310 4364 +54
- Misses 325 426 +101
Continue to review full report at Codecov.
|
xianyuanliu
removed
the
work-in-progress
| @@ -16,15 +16,17 @@ | |||
| # Dataset | |||
| # ----------------------------------------------------------------------------- | |||
| _C.DATASET = CN() | |||
| _C.DATASET.ROOT = "I:/Datasets/EgoAction/" # "/shared/tale2/Shared" | |||
| _C.DATASET.ROOT = "J:/Datasets/EgoAction/" # "/shared/tale2/Shared" | |||
kale/loaddata/videos.py
Outdated
| @@ -96,41 +126,47 @@ class VideoFrameDataset(torch.utils.data.Dataset): | |||
| might be ``jumping\0052\`` or ``sample1\`` or ``00053\``. | |||
|
|
|||
| Args: | |||
| root_path: The root path in which video folders lie. | |||
| root_path (str, Path): root path in which video folders lie. | |||
There was a problem hiding this comment.
Is Path a Python variable type?
There was a problem hiding this comment.
Changed to pathlib.Path.
kale/loaddata/videos.py
Outdated
| this is ROOT_DATA from the description above. | ||
| annotationfile_path: The .txt annotation file containing | ||
| annotationfile_path (str, Path): .txt annotation file containing |
kale/loaddata/videos.py
Outdated
| have inside of their video folders as described above. | ||
| transform: Transform pipeline that receives a list of PIL images/frames. | ||
| random_shift: Whether the frames from each segment should be taken | ||
| transform (Compose, optional): transform pipeline that receives a list of PIL images/frames. |
There was a problem hiding this comment.
torchvision.transforms.Compose
kale/loaddata/videos.py
Outdated
| image_modality (str): image modality (RGB or Optical Flow). | ||
| num_segments (int): number of segments the video should be divided into to sample frames from. | ||
| Default is 1 in image mode and 5 in feature vector mode. | ||
| frames_per_segment (int): number of frames that should |
There was a problem hiding this comment.
The lines seem do not exceed 120 characters. Please check the same problems for the other.
examples/action_dann_lightn/model.py
Outdated
| """ | ||
| Builds and returns a model and associated hyper parameters according to the config object passed. | ||
|
|
||
| Args: | ||
| cfg: A YACS config object. | ||
| dataset: A multi domain dataset consisting of source and target datasets. | ||
| num_classes: The class number of specific dataset. | ||
| dict_num_classes (dict): The dictionary of class number for specific dataset. |
There was a problem hiding this comment.
Is it better to implement this as a class (e.g., https://github.com/pykale/pykale/blob/main/kale/pipeline/domain_adapter.py#L81). We can discuss if you need.
There was a problem hiding this comment.
A function returning boolean variables is used to control image_modality and class_type. A class may be a better choice, but I have no idea about this. We can talk.
| n_verb_channel=256, | ||
| n_noun_channel=512, | ||
| dropout_keep_prob=0.5, | ||
| class_type="verb", |
There was a problem hiding this comment.
Similar to my comments above, we can implement the a class for the type of classes.
|
This pull request has been automatically marked as stale due to lack of activity. |
Description
Improve video_domain_adapter.py via organizing repeated and redundant codes into BaseAdaptTrainerVideo. This PR follows PR #291 and is followed by a large PR that will merge feature vector into action_dann_lightn example.
Status
Ready
Types of changes
docsmanually updated for new API.