Adding some baselines as transformers
closes #36 (closed)
Merge request reports
Activity
added 2 commits
added 8 commits
- 1c13eb81 - Transformed the preprocessors to transformers
- 3f9c09d7 - Created a scaler transformer
- e4649226 - Removed the transformers
- a950ce94 - Moved transformers moduled to embeddings
- 78249416 - Patched preprocessors tests
- d64f228a - Added scikit image
- 2f04b3e6 - Removed trash
- d93b03ff - Reorganized baselines
Toggle commit listadded 2 commits
mentioned in issue #36 (closed)
Closes #36 (closed)
added 9 commits
- 8f62ad63 - Cleaning up preprocessors
- 74b600e9 - Cleaning up Extractors
- ed54bf5b - Clearning up algorithms
- c2076ace - Cleaning up some unecessary entrypoints
- 1d2676e2 - Removed the old baseline mechanism
- 0a00da69 - Cleaning up the database interfaces
- 27c75ad2 - Porting some baselines
- 09ba5486 - TEsted LGBPHS transformer
- 6b891267 - Removed some unused entry_points
Toggle commit list2 2 import bob.io.image 3 3 import bob.ip.color 4 4 5 from bob.bio.base.preprocessor import Preprocessor 5 from sklearn.base import TransformerMixin, BaseEstimator 6 6 7 7 8 class Base (Preprocessor): 9 """Performs color space adaptations and data type corrections for the given 8 class Base(TransformerMixin, BaseEstimator): now I see what you are talking about on bob.pipelines#18 (closed) @tiago.pereira our preprocessors and extractors are NOT API compatible with sciki-learn transformers. we were handling one sample at a time before but now you should change the code to handle multiple samples at a time.
yeap, i've done it already
What I was trying to say in bob.pipelines#18 (closed) is just that we should well document that
.transform
should be able to handle a batch of dataThat is well known in scikit-learn https://scikit-learn.org/stable/developers/develop.html#rolling-your-own-estimator
@tiago.pereira I would like to see this issue: #26 (closed) being fixed here. The idea there is to remove the
Base
preprocessor and create several preprocessors/transofrmers that do one thing only. For example, you would have one transformer that just transformers rgb to gray. Another one for changing dype of images only. Another one just for cropping. Then, you would use the sklearn pipeline to create whatFaceCrop
does.
mentioned in commit 6d9351d7