Merge pull request #1857 from AdeelH/rel

Pre-release fixes and improvements

Author: AdeelH

.coveragerc

@@ -24,3 +24,4 @@ exclude_lines =
     @(?:abc\.)?abstractmethod
     @(?:abc\.)?abstractproperty
     @overload
+    pass

Dockerfile

@@ -1,5 +1,5 @@
-ARG UBUNTU_VERSION=22.04
 ARG CUDA_VERSION
+ARG UBUNTU_VERSION
 FROM nvidia/cuda:${CUDA_VERSION}-cudnn8-runtime-ubuntu${UBUNTU_VERSION}
 
 # wget: needed below to install conda

docker/build

@@ -13,7 +13,7 @@ function usage() {
 Build Docker images.
 
 Options:
---arm64 will build image for arm64 architecture
+--arm64   Build image for arm64 architecture.
 "
 }
 
@@ -27,13 +27,28 @@ then
 
     PLATFORM="amd64"
     IMAGE_EXT=""
-    if [ "${1:-}" = "--arm64" ]
-    then
-        PLATFORM="arm64"
-        IMAGE_EXT="-arm64"
-    fi
+    CUDA_VERSION="12.1.1"
+    UBUNTU_VERSION="22.04"
+
+    while [[ $# -gt 0 ]]
+    do
+        case "$1" in
+            --arm64)
+                PLATFORM="arm64"
+                IMAGE_EXT="-arm64"
+                shift
+                ;;
+            *)
+                echo "Unknown option: $1"
+                usage
+                exit 1
+                ;;
+        esac
+    done
 
     DOCKER_BUILDKIT=1 docker build \
-        --platform linux/${PLATFORM} --build-arg CUDA_VERSION="11.7.1" \
+        --platform linux/${PLATFORM} \
+        --build-arg CUDA_VERSION="${CUDA_VERSION}" \
+        --build-arg UBUNTU_VERSION="${UBUNTU_VERSION}" \
         -t raster-vision-pytorch${IMAGE_EXT} -f Dockerfile .
 fi

docker/run

@@ -107,7 +107,7 @@ do
         ;;
         --jupyter-lab)
         find_free_port_in_range 8888 9999
-        JUPYTER="-v ${RASTER_VISION_NOTEBOOK_DIR}:/opt/notebooks -v ${HOME}/.jupyter:/root/.jupyter:ro -p $FREE_PORT:$FREE_PORT"
+        JUPYTER="-v ${RASTER_VISION_NOTEBOOK_DIR}:/opt/notebooks -v ${HOME}/.jupyter:/root/.jupyter -p $FREE_PORT:$FREE_PORT"
         # run jupyter lab in the background
         CMD=(/bin/bash -c "jupyter lab --ip 0.0.0.0 --port $FREE_PORT --no-browser --allow-root --notebook-dir=/opt/notebooks & bash")
         echo "Starting Jupyter Lab server at 0.0.0.0:$FREE_PORT. This may take a few seconds."

docs/release.rst

@@ -9,32 +9,44 @@ Minor or Major Version Release
 ------------------------------
 
 #.  It's a good idea to update any major dependencies before the release.
-#.  Update the docs if needed. See the `docs README <{{ repo }}/docs/README.md>`__ for instructions.
-#.  Checkout the ``master`` branch, re-build the docker image (``docker/build``), and push it to ECR (``docker/ecr_publish``).
-#.  Execute all `tutorial notebooks <{{ repo }}/docs/usage/tutorials/>`__ and make sure they work correctly. Do not commit output changes unless code behavior has changed.
-#.  Run all :ref:`rv examples` and check that evaluation metrics are close to the scores from the last release. (For each example, there should be a link to a JSON file with the evaluation metrics from the last release.) This stage often uncovers bugs, and is the most time consuming part of the release process. There is a `script <{{ repo_examples }}/test.py>`__ to help run the examples and collect their outputs. See the associated `README <{{ repo_examples }}/README.md>`__ for details.
-#.  Collect all model bundles, and check that they work with the ``predict`` command and sanity check output in QGIS.
-#.  Update the :ref:`model zoo` by uploading model bundles and sample images to the right place on S3. If you use the ``collect`` command (`described here <{{ repo_examples }}/README.md>`__), you should be able to sync the ``collect_dir`` to ``s3://azavea-research-public-data/raster-vision/examples/model-zoo-<version>``.
-#.  Update the notebooks that use models from the model zoo so that they use the latest version and re-run.
-#.  Update `tiny_spacenet.py <{{ repo_examples }}/tiny_spacenet.py>`__ if needed and ensure the line numbers in every ``literalinclude`` of that file are correct. Tip: you can find all instances by searching the repo using the regex: ``\.\. literalinclude:: .+tiny_spacenet\.py$``.
-#.  Test :ref:`setup` and :ref:`quickstart` instructions and make sure they work.
-#.  Test examples from :ref:`pipelines plugins`.
+#.  Test examples:
 
-    .. code-block:: console
+    #.  Checkout the ``master`` branch, re-build the docker image (``docker/build``), and push it to ECR (``docker/ecr_publish``).
+    #. Follow the instructions in `this README <{{ repo_examples }}/README.md>`__ to do the following:
+        
+       #.  Run all :ref:`rv examples` and check that evaluation metrics are close to the scores from the last release. (For each example, there should be a link to a JSON file with the evaluation metrics from the last release.) This stage often uncovers bugs, and is the most time consuming part of the release process.
+       #.  Collect all model bundles, and check that they work with the ``predict`` command and sanity check output in QGIS.
+       #.  Update the :ref:`model zoo` by uploading model bundles and sample images to the right place on S3. If you use the ``collect`` command (`see <{{ repo_examples }}/README.md>`__), you should be able to sync the ``collect_dir`` to ``s3://azavea-research-public-data/raster-vision/examples/model-zoo-<version>``.
+       #. Screenshot the outputs of the ``compare`` command (for each example) and include them in the PR described below.
 
-        rastervision run inprocess rastervision.pipeline_example_plugin1.config1 -a root_uri /opt/data/pipeline-example/1/ --splits 2
-        rastervision run inprocess rastervision.pipeline_example_plugin1.config2 -a root_uri /opt/data/pipeline-example/2/ --splits 2
-        rastervision run inprocess rastervision.pipeline_example_plugin2.config3 -a root_uri /opt/data/pipeline-example/3/ --splits 2
+#.  Test notebooks:
 
-#.  Test examples from :ref:`bootstrap`.
+    #.  Update the `tutorial notebooks <{{ repo }}/docs/usage/tutorials/>`__ that use models from the model zoo so that they use the latest version.
+    #.  Execute all `tutorial notebooks <{{ repo }}/docs/usage/tutorials/>`__ and make sure they work correctly. Do not commit output changes unless code behavior has changed.
 
-    .. code-block:: console
+#. Test/update docs:
+
+   #.  Update the docs if needed. See the `docs README <{{ repo }}/docs/README.md>`__ for instructions.
+   #.  Update `tiny_spacenet.py <{{ repo_examples }}/tiny_spacenet.py>`__ if needed and ensure the line numbers in every ``literalinclude`` of that file are correct. Tip: you can find all instances by searching the repo using the regex: ``\.\. literalinclude:: .+tiny_spacenet\.py$``.
+   #.  Test :ref:`setup` and :ref:`quickstart` instructions and make sure they work.
+   #.  Test examples from :ref:`pipelines plugins`.
+
+        .. code-block:: console
+
+            rastervision run inprocess rastervision.pipeline_example_plugin1.config1 -a root_uri /opt/data/pipeline-example/1/ --splits 2
+            rastervision run inprocess rastervision.pipeline_example_plugin1.config2 -a root_uri /opt/data/pipeline-example/2/ --splits 2
+            rastervision run inprocess rastervision.pipeline_example_plugin2.config3 -a root_uri /opt/data/pipeline-example/3/ --splits 2
+
+   #.  Test examples from :ref:`bootstrap`.
+
+       .. code-block:: console
+
+           cookiecutter /opt/src/cookiecutter_template
 
-        cookiecutter /opt/src/cookiecutter_template
+   #.  Update the `the changelog <{{ repo }}/docs/changelog.rst>`__, and point out API changes.
+   #.  Fix any broken badges on the GitHub repo readme.
 
-#.  Update the `the changelog <{{ repo }}/docs/changelog.rst>`__, and point out API changes.
-#.  Fix any broken badges on the GitHub repo readme.
-#.  Update the version number. This occurs in several places, so it's best to do this with a find and replace over the entire repo.
+#.  Update the version number. This occurs in several places, so it's best to do this with a find-and-replace over the entire repo.
 #.  Make a PR to the ``master`` branch with the preceding updates. In the PR, there should be a link to preview the docs. Check that they are building and look correct.
 #.  Make a git branch with the version as the name, and push to GitHub.
 #.  Ensure that the docs are building correctly for the new version branch on `readthedocs <https://readthedocs.org/projects/raster-vision/>`_. You will need to have admin access on your RTD account. Once the branch is building successfully, Under *Versions -> Activate a Version*, you can activate the version to add it to the sidebar of the docs for the latest version. (This might require manually triggering a rebuild of the docs.) Then, under *Admin -> Advanced Settings*, change the default version to the new version.

docs/usage/tutorials/pred_and_eval_ss.ipynb

@@ -43,7 +43,7 @@
    },
    "source": [
     "Load a :class:`Learner` with a trained model from bundle -- :meth:`.Learner.from_model_bundle`\n",
-    "---------------------------------------------------------------------------------------------"
+    "----------------------------------------------------------------------------------------------"
    ]
   },
   {

rastervision_core/rastervision/core/box.py

@@ -1,4 +1,5 @@
-from typing import TYPE_CHECKING, Callable, Dict, Union, Tuple, Optional, List
+from typing import (TYPE_CHECKING, Callable, Dict, List, Optional, Sequence,
+                    Tuple, Union)
 from typing_extensions import Literal
 from pydantic import PositiveInt as PosInt, conint
 import math
@@ -484,3 +485,22 @@ def within_aoi(window: 'Box', aoi_polygons: List[Polygon]) -> bool:
             if w.within(polygon):
                 return True
         return False
+
+    def __contains__(self, query: Union['Box', Sequence]) -> bool:
+        """Check if box or point is contained within this box.
+
+        Args:
+            query: Box or single point (x, y).
+
+        Raises:
+            NotImplementedError: if query is not a Box or tuple/list.
+        """
+        if isinstance(query, Box):
+            ymin, xmin, ymax, xmax = query
+            return (ymin >= self.ymin and xmin >= self.xmin
+                    and ymax <= self.ymax and xmax <= self.xmax)
+        elif isinstance(query, (tuple, list)):
+            x, y = query
+            return self.xmin <= x <= self.xmax and self.ymin <= y <= self.ymax
+        else:
+            raise NotImplementedError()

rastervision_core/rastervision/core/data/label/chip_classification_labels.py

@@ -118,7 +118,7 @@ def get_cell_scores(self, cell: Box) -> Optional[Sequence[float]]:
         """
         result = self.cell_to_label.get(cell)
         if result is not None:
-            return result.score
+            return result.scores
         else:
             return None
 

rastervision_core/rastervision/core/data/label/object_detection_labels.py

@@ -163,7 +163,7 @@ def get_class_ids(self) -> np.ndarray:
     def __len__(self) -> int:
         return self.boxlist.get().shape[0]
 
-    def __str__(self) -> str:
+    def __str__(self) -> str:  # prama: no cover
         return str(self.boxlist.get())
 
     def to_boxlist(self) -> NpBoxList:

rastervision_core/rastervision/core/data/label/semantic_segmentation_labels.py

@@ -1,5 +1,4 @@
-from typing import (TYPE_CHECKING, Any, Iterable, List, Optional, Sequence,
-                    Union)
+from typing import (TYPE_CHECKING, Any, Iterable, List, Optional, Sequence)
 from abc import abstractmethod
 
 import numpy as np
@@ -23,8 +22,8 @@ def __init__(self, extent: Box, num_classes: int, dtype: np.dtype):
         """Constructor.
 
         Args:
-            extent (Box): The extent of the region to which
-                the labels belong, in global coordinates.
+            extent (Box): The extent of the region to which the labels belong,
+                in global coordinates.
             num_classes (int): Number of classes.
         """
         self.extent = extent
@@ -143,9 +142,8 @@ def transform_shape(x, y, z=None):
             del self[window]
 
     @classmethod
-    def make_empty(cls, extent: Box, num_classes: int, smooth: bool = False
-                   ) -> Union['SemanticSegmentationDiscreteLabels',
-                              'SemanticSegmentationSmoothLabels']:
+    def make_empty(cls, extent: Box, num_classes: int,
+                   smooth: bool = False) -> 'SemanticSegmentationLabels':
         """Instantiate an empty instance.
 
         Args:
@@ -157,8 +155,7 @@ def make_empty(cls, extent: Box, num_classes: int, smooth: bool = False
                 SemanticSegmentationDiscreteLabels object. Defaults to False.
 
         Returns:
-            Union[SemanticSegmentationDiscreteLabels,
-            SemanticSegmentationSmoothLabels]: If smooth=True, returns a
+            SemanticSegmentationLabels: If smooth=True, returns a
                 SemanticSegmentationSmoothLabels. Otherwise, a
                 SemanticSegmentationDiscreteLabels.
 
@@ -174,15 +171,14 @@ def make_empty(cls, extent: Box, num_classes: int, smooth: bool = False
                 extent=extent, num_classes=num_classes)
 
     @classmethod
-    def from_predictions(cls,
-                         windows: Iterable['Box'],
-                         predictions: Iterable[Any],
-                         extent: Box,
-                         num_classes: int,
-                         smooth: bool = False,
-                         crop_sz: Optional[int] = None
-                         ) -> Union['SemanticSegmentationDiscreteLabels',
-                                    'SemanticSegmentationSmoothLabels']:
+    def from_predictions(
+            cls,
+            windows: Iterable['Box'],
+            predictions: Iterable[Any],
+            extent: Box,
+            num_classes: int,
+            smooth: bool = False,
+            crop_sz: Optional[int] = None) -> 'SemanticSegmentationLabels':
         """Instantiate from windows and their corresponding predictions.
 
         Args:
@@ -202,8 +198,7 @@ def from_predictions(cls,
                 windows. Defaults to None.
 
         Returns:
-            Union[SemanticSegmentationDiscreteLabels,
-            SemanticSegmentationSmoothLabels]: If smooth=True, returns a
+            SemanticSegmentationLabels: If smooth=True, returns a
                 SemanticSegmentationSmoothLabels. Otherwise, a
                 SemanticSegmentationDiscreteLabels.
         """
@@ -349,8 +344,7 @@ def from_predictions(cls,
                          extent: Box,
                          num_classes: int,
                          crop_sz: Optional[int] = None
-                         ) -> Union['SemanticSegmentationDiscreteLabels',
-                                    'SemanticSegmentationSmoothLabels']:
+                         ) -> 'SemanticSegmentationDiscreteLabels':
         labels = cls.make_empty(extent, num_classes)
         labels.add_predictions(windows, predictions, crop_sz=crop_sz)
         return labels
@@ -522,8 +516,7 @@ def from_predictions(cls,
                          extent: Box,
                          num_classes: int,
                          crop_sz: Optional[int] = None
-                         ) -> Union['SemanticSegmentationDiscreteLabels',
-                                    'SemanticSegmentationSmoothLabels']:
+                         ) -> 'SemanticSegmentationSmoothLabels':
         labels = cls.make_empty(extent, num_classes)
         labels.add_predictions(windows, predictions, crop_sz=crop_sz)
         return labels

rastervision_core/rastervision/core/data/label_source/semantic_segmentation_label_source.py

@@ -7,24 +7,12 @@
 from rastervision.core.data.label import SemanticSegmentationLabels
 from rastervision.core.data.label_source.label_source import LabelSource
 from rastervision.core.data.raster_source import RasterSource
+from rastervision.core.data.utils import pad_to_window_size
 
 if TYPE_CHECKING:
     from rastervision.core.data import CRSTransformer
 
 
-def fill_edge(label_arr: np.ndarray, window: Box, extent: Box,
-              fill_value: int) -> np.ndarray:
-    """If window goes over the edge of the extent, buffer with fill_value."""
-    if window.ymax <= extent.ymax and window.xmax <= extent.xmax:
-        return label_arr
-
-    x = np.full(window.size, fill_value)
-    ylim = extent.ymax - window.ymin
-    xlim = extent.xmax - window.xmin
-    x[0:ylim, 0:xlim] = label_arr[0:ylim, 0:xlim]
-    return x
-
-
 class SemanticSegmentationLabelSource(LabelSource):
     """A read-only label source for semantic segmentation."""
 
@@ -101,8 +89,9 @@ def get_label_arr(self, window: Optional[Box] = None) -> np.ndarray:
         of the null class as defined by the class_config.
 
         Args:
-            window (Optional[Box], optional): Window to get labels for. If
-            None, returns a label array covering the full extent of the scene.
+            window (Optional[Box], optional): Window (in pixel coords) to get
+                labels for. If None, returns a label array covering the full
+                extent of the scene.
 
         Returns:
             np.ndarray: Label array.
@@ -113,8 +102,10 @@ def get_label_arr(self, window: Optional[Box] = None) -> np.ndarray:
         label_arr = self.raster_source.get_chip(window)
         if label_arr.ndim == 3:
             label_arr = np.squeeze(label_arr, axis=2)
-        label_arr = fill_edge(label_arr, window, self.extent,
-                              self.class_config.null_class_id)
+        h, w = label_arr.shape
+        if h < window.height or w < window.width:
+            label_arr = pad_to_window_size(label_arr, window, self.extent,
+                                           self.class_config.null_class_id)
         return label_arr
 
     @property

rastervision_core/rastervision/core/data/label_store/semantic_segmentation_label_store.py

@@ -17,6 +17,7 @@
 from rastervision.core.data.label_source import SemanticSegmentationLabelSource
 from rastervision.core.data.raster_transformer import RGBClassTransformer
 from rastervision.core.data.raster_source import RasterioSource
+from rastervision.core.data.utils import write_window
 
 if TYPE_CHECKING:
     from rastervision.core.data import (VectorOutputConfig,
@@ -266,7 +267,8 @@ def write_smooth_raster_output(
                         score_arr = self._scores_to_uint8(score_arr)
                     else:
                         score_arr = score_arr.astype(dtype)
-                    self._write_array(ds, window, score_arr)
+                    score_arr = score_arr.transpose(1, 2, 0)
+                    write_window(ds, score_arr, window)
 
         # save pixel hits too
         np.save(hits_path, labels.pixel_hits)
@@ -291,8 +293,7 @@ def write_discrete_raster_output(
                     if self.class_transformer is not None:
                         label_arr = self.class_transformer.class_to_rgb(
                             label_arr)
-                        label_arr = label_arr.transpose(2, 0, 1)
-                    self._write_array(ds, window, label_arr)
+                    write_window(ds, label_arr, window)
 
     def write_vector_outputs(self, labels: SemanticSegmentationLabels,
                              vector_output_dir: str) -> None:
@@ -314,18 +315,6 @@ def write_vector_outputs(self, labels: SemanticSegmentationLabels,
                 out_uri = vo.get_uri(vector_output_dir, self.class_config)
                 json_to_file(geojson, out_uri)
 
-    def _write_array(self, dataset: rio.DatasetReader, window: Box,
-                     arr: np.ndarray) -> None:
-        """Write array out to a rasterio dataset. Array must be of shape
-        (C, H, W).
-        """
-        rio_window = window.rasterio_format()
-        if len(arr.shape) == 2:
-            dataset.write_band(1, arr, window=rio_window)
-        else:
-            for i, band in enumerate(arr, start=1):
-                dataset.write_band(i, band, window=rio_window)
-
     def _clip_to_extent(self,
                         extent: Box,
                         window: Box,