Feat (explorer view): Simplify UI (#237)

* feat: add doc_url filed to metrics

* style: remove commented codes

* feat: point to main doc page if attribute does not exist

* feat: do not show any link if doc_url is empty

* fix: change default doc_url value

* feat: update docs and example_metric

docs/docs/metrics/write-your-own.md

@@ -47,6 +47,7 @@ to the awesome paper that proposed the method.
 Or use math to better explain such method:
 $$h_{\lambda}(x) = \frac{1}{x^\intercal x}$$
 """,
+            doc_url='link/to/documentation', # This is optional, if a link is given, it can be accessed from the app
             metric_type=MetricType.HEURISTIC,
             data_type=DataType.IMAGE,
             annotation_type=[AnnotationType.OBJECT.BOUNDING_BOX, AnnotationType.OBJECT.ROTATABLE_BOUNDING_BOX, AnnotationType.OBJECT.POLYGON],

src/encord_active/app/data_quality/sub_pages/explorer.py

@@ -117,9 +117,10 @@ def build(self, selected_df: DataFrame[MetricSchema], metric_scope: MetricScope)
         if not selected_metric:
             return
 
-        st.markdown(f"# {self.title}")
-        st.markdown(f"## {selected_metric.meta.title}")
-        st.markdown(selected_metric.meta.long_description)
+        if selected_metric.meta.doc_url is None:
+            st.markdown(f"### {selected_metric.meta.title}")
+        else:
+            st.markdown(f"### [{selected_metric.meta.title}]({selected_metric.meta.doc_url})")
 
         if selected_df.empty:
             return

src/encord_active/lib/metrics/example.py

@@ -26,6 +26,7 @@ def __init__(self):
 Or use math to better explain such method:
 $$h_{\lambda}(x) = \frac{1}{x^\intercal x}$$
 """,
+            doc_url="link/to/documentation",  # This is optional, if a link is given, it can be accessed from the app
             metric_type=MetricType.HEURISTIC,
             data_type=DataType.IMAGE,
             annotation_type=[

src/encord_active/lib/metrics/geometric/annotation_duplicates.py

@@ -23,6 +23,7 @@ def __init__(self, threshold: float = 0.6):
             long_description=r"""Ranks annotations by how likely they are to represent the same object.
 > [Jaccard similarity coefficient](https://en.wikipedia.org/wiki/Jaccard_index)
 is used to measure closeness of two annotations.""",
+            doc_url="https://encord-active-docs.web.app/metrics/geometric#annotation-duplicates",
             metric_type=MetricType.GEOMETRIC,
             data_type=DataType.IMAGE,
             annotation_type=[

src/encord_active/lib/metrics/geometric/hu_static.py

@@ -40,6 +40,7 @@ def __init__(self):
             long_description=r"""Computes the Euclidean distance between the polygons'
     [Hu moments](https://en.wikipedia.org/wiki/Image_moment) for each class and
     the prototypical class moments.""",
+            doc_url="https://encord-active-docs.web.app/metrics/geometric#shape-outlier-detection",
             metric_type=MetricType.GEOMETRIC,
             data_type=DataType.IMAGE,
             annotation_type=[AnnotationType.OBJECT.POLYGON],

src/encord_active/lib/metrics/geometric/hu_temporal.py

@@ -53,6 +53,7 @@ def __init__(self):
             long_description=r"""Ranks objects by how similar they are to their instances in previous frames
 based on [Hu moments](https://en.wikipedia.org/wiki/Image_moment). The more an object's shape changes,
 the lower its score will be.""",
+            doc_url="https://encord-active-docs.web.app/metrics/geometric#polygon-shape-similarity",
             metric_type=MetricType.GEOMETRIC,
             data_type=DataType.SEQUENCE,
             annotation_type=[AnnotationType.OBJECT.POLYGON],

src/encord_active/lib/metrics/geometric/image_border_closeness.py

@@ -19,6 +19,7 @@ def __init__(self):
             title="Annotation closeness to image borders",
             short_description="Ranks annotations by how close they are to image borders.",
             long_description=r"""This metric ranks annotations by how close they are to image borders.""",
+            doc_url="https://encord-active-docs.web.app/metrics/geometric#polygon-shape-similarity",
             metric_type=MetricType.GEOMETRIC,
             data_type=DataType.IMAGE,
             annotation_type=[

src/encord_active/lib/metrics/geometric/object_size.py

@@ -35,6 +35,7 @@ def __init__(self):
             title="Object Area - Relative",
             short_description="Computes object area as a percentage of total image area.",
             long_description=r"""Computes object area as a percentage of total image area.""",
+            doc_url="https://encord-active-docs.web.app/metrics/geometric#polygon-shape-similarity",
             metric_type=MetricType.GEOMETRIC,
             data_type=DataType.IMAGE,
             annotation_type=[
@@ -68,6 +69,7 @@ def __init__(self):
             title="Frame object density",
             short_description="Computes the percentage of image area that's occupied by objects.",
             long_description=r"""Computes the percentage of image area that's occupied by objects.""",
+            doc_url="https://encord-active-docs.web.app/metrics/geometric#polygon-shape-similarity",
             metric_type=MetricType.GEOMETRIC,
             data_type=DataType.IMAGE,
             annotation_type=[
@@ -111,6 +113,7 @@ def __init__(self):
             title="Object Area - Absolute",
             short_description="Computes object area in amount of pixels",
             long_description=r"""Computes object area in amount of pixels.""",
+            doc_url="https://encord-active-docs.web.app/metrics/geometric#object-area---absolute",
             metric_type=MetricType.GEOMETRIC,
             data_type=DataType.IMAGE,
             annotation_type=[
@@ -151,6 +154,7 @@ def __init__(self):
             title="Object Aspect Ratio",
             short_description="Computes aspect ratios of objects",
             long_description=r"""Computes aspect ratios ($width/height$) of objects.""",
+            doc_url="https://encord-active-docs.web.app/metrics/geometric#object-aspect-ratio",
             metric_type=MetricType.GEOMETRIC,
             data_type=DataType.IMAGE,
             annotation_type=[

src/encord_active/lib/metrics/geometric/occlusion_detection_video.py

@@ -28,6 +28,7 @@ def __init__(
             short_description="Tracks objects and detect outliers",
             long_description=r"""This metric collects information related to object size and aspect ratio for each track
  and find outliers among them.""",
+            doc_url="https://encord-active-docs.web.app/metrics/geometric#detect-occlusion-in-video",
             metric_type=MetricType.GEOMETRIC,
             data_type=DataType.SEQUENCE,
             annotation_type=[AnnotationType.OBJECT.BOUNDING_BOX, AnnotationType.OBJECT.ROTATABLE_BOUNDING_BOX],

src/encord_active/lib/metrics/heuristic/high_iou_changing_classes.py

@@ -54,6 +54,7 @@ def __init__(self, threshold: float = 0.8):
 `Cat:2` will be flagged as potentially having a broken track, because track ids `1` and `2` doesn't match.
 
 """,
+            doc_url="https://encord-active-docs.web.app/metrics/heuristic#inconsistent-object-classification-and-track-ids",
             metric_type=MetricType.HEURISTIC,
             data_type=DataType.SEQUENCE,
             annotation_type=[

src/encord_active/lib/metrics/heuristic/img_features.py

@@ -24,6 +24,7 @@ def __init__(self):
 
 Contrast is computed as the standard deviation of the pixel values.
 """,
+            doc_url="https://encord-active-docs.web.app/metrics/heuristic#contrast",
             metric_type=MetricType.HEURISTIC,
             data_type=DataType.IMAGE,
             annotation_type=AnnotationType.NONE,
@@ -49,6 +50,7 @@ def __init__(
                 short_description=f"Ranks images by how {color_name.lower()} the average value of the image is.",
                 long_description=f"""Ranks images by how {color_name.lower()} the average value of the
                     image is.""",
+                doc_url=f"https://encord-active-docs.web.app/metrics/heuristic/#{color_name.lower()}-values",
                 metric_type=MetricType.HEURISTIC,
                 data_type=DataType.IMAGE,
                 annotation_type=AnnotationType.NONE,
@@ -138,6 +140,7 @@ def __init__(self):
 
 Brightness is computed as the average (normalized) pixel value across each image.
 """,
+            doc_url="https://encord-active-docs.web.app/metrics/heuristic#brightness",
             metric_type=MetricType.HEURISTIC,
             data_type=DataType.IMAGE,
             annotation_type=AnnotationType.NONE,
@@ -163,6 +166,7 @@ def __init__(self):
 score = cv2.Laplacian(image, cv2.CV_64F).var()
 ```
 """,
+            doc_url="https://encord-active-docs.web.app/metrics/heuristic#sharpness",
             metric_type=MetricType.HEURISTIC,
             data_type=DataType.IMAGE,
             annotation_type=AnnotationType.NONE,
@@ -188,6 +192,7 @@ def __init__(self):
 score = 1 - cv2.Laplacian(image, cv2.CV_64F).var()
 ```
 """,
+            doc_url="https://encord-active-docs.web.app/metrics/heuristic#blur",
             metric_type=MetricType.HEURISTIC,
             data_type=DataType.IMAGE,
             annotation_type=AnnotationType.NONE,
@@ -207,6 +212,7 @@ def __init__(self):
 
 Aspect ratio is computed as the ratio of image width to image height ($\frac{width}{height}$).
 """,
+            doc_url="https://encord-active-docs.web.app/metrics/heuristic#aspect-ratio",
             metric_type=MetricType.HEURISTIC,
             data_type=DataType.IMAGE,
             annotation_type=AnnotationType.NONE,
@@ -231,6 +237,7 @@ def __init__(self):
 
 Area is computed as the product of image width and image height ($width \times height$).
     """,
+            doc_url="https://encord-active-docs.web.app/metrics/heuristic#area",
             metric_type=MetricType.HEURISTIC,
             data_type=DataType.IMAGE,
             annotation_type=AnnotationType.NONE,

src/encord_active/lib/metrics/heuristic/missing_objects_and_wrong_tracks.py

@@ -80,6 +80,7 @@ def __init__(self, threshold: float = 0.5):
 ```
 `CAT:2` will be marked as potentially having a wrong track id.
 """,
+            doc_url="https://encord-active-docs.web.app/metrics/heuristic#missing-objects-and-broken-tracks",
             metric_type=MetricType.HEURISTIC,
             data_type=DataType.SEQUENCE,
             annotation_type=[

src/encord_active/lib/metrics/heuristic/object_counting.py

@@ -14,6 +14,7 @@ def __init__(self):
             title="Object Count",
             short_description="Counts number of objects in the image",
             long_description=r"""Counts number of objects in the image.""",
+            doc_url="https://encord-active-docs.web.app/metrics/heuristic#object-count",
             metric_type=MetricType.HEURISTIC,
             data_type=DataType.IMAGE,
             annotation_type=AnnotationType.ALL,

src/encord_active/lib/metrics/heuristic/random.py

@@ -16,6 +16,7 @@ def __init__(self):
             title="Random Values on Images",
             short_description="Assigns a random value between 0 and 1 to images",
             long_description="Uses a uniform distribution to generate a value between 0 and 1 to each image",
+            doc_url="https://encord-active-docs.web.app/metrics/heuristic#random-values-on-images",
             metric_type=MetricType.HEURISTIC,
             data_type=DataType.IMAGE,
             annotation_type=[],
@@ -34,6 +35,7 @@ def __init__(self):
             title="Random Values on Objects",
             short_description="Assigns a random value between 0 and 1 to objects",
             long_description="Uses a uniform distribution to generate a value between 0 and 1 to each object",
+            doc_url="https://encord-active-docs.web.app/metrics/heuristic#random-values-on-objects",
             metric_type=MetricType.HEURISTIC,
             data_type=DataType.IMAGE,
             annotation_type=[

src/encord_active/lib/metrics/metric.py

@@ -65,6 +65,7 @@ class MetricMetadata(BaseModel):
     data_type: DataType
     annotation_type: List[AnnotationTypeUnion]
     embedding_type: Optional[EmbeddingType] = None
+    doc_url: Optional[str] = None
     stats: StatsMetadata
 
     def get_unique_name(self):
@@ -82,6 +83,7 @@ def __init__(
         data_type: DataType,
         annotation_type: List[Union[ObjectShape, ClassificationType]],
         embedding_type: Optional[EmbeddingType] = None,
+        doc_url: Optional[str] = None,
     ):
         self.metadata = MetricMetadata(
             title=title,
@@ -91,6 +93,7 @@ def __init__(
             data_type=data_type,
             annotation_type=annotation_type,
             embedding_type=embedding_type,
+            doc_url=doc_url,
             stats=StatsMetadata(),
         )
 
@@ -109,6 +112,7 @@ def __init__(
         data_type: DataType,
         annotation_type: List[Union[ObjectShape, ClassificationType]] = [],
         embedding_type: Optional[EmbeddingType] = None,
+        doc_url: Optional[str] = None,
     ):
         self.metadata = MetricMetadata(
             title=title,
@@ -118,6 +122,7 @@ def __init__(
             data_type=data_type,
             annotation_type=annotation_type,
             embedding_type=embedding_type,
+            doc_url=doc_url,
             stats=StatsMetadata(),
         )
 

src/encord_active/lib/metrics/semantic/image_singularity.py

@@ -37,6 +37,7 @@ def __init__(self, near_duplicate_threshold=0.97):
 - **To delete duplicate images:** You can set the quality filter to cover only zero values (that ends up with all the duplicate images), then use bulk tagging (e.g., with a tag like `Duplicate`) to tag all images.
 - **To mark duplicate images:** Near duplicate images are shown side by side. Navigate through these images and mark whichever is of interest to you.
 """,
+            doc_url="https://encord-active-docs.web.app/metrics/semantic#image-singularity",
             metric_type=MetricType.SEMANTIC,
             data_type=DataType.IMAGE,
             embedding_type=EmbeddingType.CLASSIFICATION,

src/encord_active/lib/metrics/semantic/img_classification_quality.py

@@ -57,6 +57,7 @@ def __init__(self, num_nearest_neighbors: int = 10, certainty_ratio: float = 0.6
             long_description=r"""This metric creates embeddings from images. Then, these embeddings are used to build
     nearest neighbor graph. Similar embeddings' classifications are compared against each other.
         """,
+            doc_url="https://encord-active-docs.web.app/metrics/semantic#image-level-annotation-quality",
             metric_type=MetricType.SEMANTIC,
             data_type=DataType.IMAGE,
             annotation_type=[AnnotationType.CLASSIFICATION.RADIO],

src/encord_active/lib/metrics/semantic/img_object_quality.py

@@ -43,6 +43,7 @@ def __init__(self, num_nearest_neighbors: int = 10, certainty_ratio: float = 0.6
     and an embedding for each bounding box is extracted. Then, these embeddings are compared
     with their neighbors. If the neighbors are annotated differently, a low score is given to it.
     """,
+            doc_url="https://encord-active-docs.web.app/metrics/semantic#object-annotation-quality",
             metric_type=MetricType.GEOMETRIC,
             data_type=DataType.IMAGE,
             annotation_type=[

src/encord_active/lib/metrics/utils.py

@@ -130,6 +130,7 @@ def load_metric_metadata(meta_pth) -> MetricMetadata:
             metric_type=old_meta.get("metric_type", old_meta.get("index_type")),
             embedding_type=old_meta.get("embedding_type", None),
             annotation_type=annotation_type,
+            doc_url=old_meta.get("doc_url", None),
             stats=stats,
         )
         with meta_pth.open("w") as f: