Feature/semseg misclassification #628
Conversation
Codecov Report
@@ Coverage Diff @@
## main #628 +/- ##
==========================================
- Coverage 84.39% 83.76% -0.64%
==========================================
Files 164 165 +1
Lines 12964 12959 -5
==========================================
- Hits 10941 10855 -86
- Misses 2023 2104 +81
... and 11 files with indirect coverage changes 📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more |
| accuracy: Optional[float] = None | ||
| dominant_mislabel_class: Optional[int] = None | ||
| dominant_mislabel_class_percent: Optional[float] = None |
There was a problem hiding this comment.
if these are all present when we instantiate the object, we don't need them to be optional
| @@ -15,6 +15,20 @@ class SemSegCols(str, Enum): | |||
| meta = "meta" | |||
|
|
|||
|
|
|||
| class ClassificationData(BaseModel): | |||
There was a problem hiding this comment.
Maybe ClassificationErrorData
| @@ -15,6 +15,20 @@ class SemSegCols(str, Enum): | |||
| meta = "meta" | |||
|
|
|||
|
|
|||
| class ClassificationData(BaseModel): | |||
| """ | |||
| accuracy: float the mean accuracy per pixel | |||
There was a problem hiding this comment.
what does "mean accuracy per pixel" mean? specifically the word "mean" in this context
| @@ -30,7 +44,7 @@ class IoUType(str, Enum): | |||
| class IouData(BaseModel): | |||
| iou: float | |||
| iou_per_class: List[float] | |||
| area_per_class: List[int] | |||
| area_per_class: List[float] | |||
There was a problem hiding this comment.
if this is a float are the other area values a float? if so we should update the fn typing to return float not int
| @@ -62,7 +76,7 @@ class Contour(BaseModel): | |||
| class Polygon(BaseModel): | |||
| uuid: str # UUID4 | |||
| label_idx: int | |||
| misclassified_class_label: Optional[int] = None | |||
| classification_data: ClassificationData = ClassificationData() | |||
There was a problem hiding this comment.
just have this be optional and default to None
| add_classification_error_to_polygons_batch( | ||
| self.pred_masks, gold_polygons_batch, n_classes | ||
| ) | ||
| add_classification_error_to_polygons_batch( | ||
| self.gold_masks, pred_polygons_batch, n_classes | ||
| ) |
There was a problem hiding this comment.
why are we doing it for both now?
There was a problem hiding this comment.
We want both to have classification accuracy and errors
|
To fix typing issue in 3.8 and 3.9 https://stackoverflow.com/questions/76313592/import-langchain-error-typeerror-issubclass-arg-1-must-be-a-class |
Adds misclassification to semseg polygons