Merge pull request #31 from hasty-ai/kp-add-automated-labeling

Add automated labeling endpoint support

Author: proskudin

hasty/__init__.py

@@ -19,7 +19,7 @@ def int_or_str(value):
         return value
 
 
-__version__ = '0.3.5'
+__version__ = '0.3.6'
 VERSION = tuple(map(int_or_str, __version__.split('.')))
 
 __all__ = [

hasty/automated_labeling.py

@@ -0,0 +1,117 @@
+from collections import OrderedDict
+from time import time
+from typing import Optional
+
+from .constants import WAIT_INTERVAL_SEC
+from .hasty_object import HastyObject
+from .requester import Requester
+
+
+class AutomatedLabelingJob(HastyObject):
+    endpoint = '/v1/projects/{project_id}/automated_labeling'
+    endpoint_job_id = '/v1/projects/{project_id}/automated_labeling/{job_id}'
+
+    def __repr__(self):
+        return self.get__repr__(OrderedDict({"id": self._id,
+                                             "status": self._status,
+                                             "progress": self._progress}))
+
+    @property
+    def id(self):
+        """
+        :type: string
+        """
+        return self._id
+
+    @property
+    def status(self):
+        """
+        :type: string
+        """
+        return self._status
+
+    @property
+    def project_id(self):
+        """
+        :type: string
+        """
+        return self._project_id
+
+    @property
+    def progress(self):
+        """
+        :type: float
+        """
+        return self._progress
+
+    @property
+    def started_on(self):
+        """
+        :type: string
+        """
+        return self._started_on
+
+    @property
+    def completed_on(self):
+        """
+        :type: string
+        """
+        return self._completed_on
+
+    @property
+    def experiment_id(self):
+        """
+        :type: string
+        """
+        return self._experiment_id
+
+    def _init_properties(self):
+        self._job_id = None
+        self._status = None
+        self._project_id = None
+        self._started_on = None
+        self._completed_on = None
+        self._last_check = None
+        self._experiment_id = None
+        self._progress = None
+
+    def _set_prop_values(self, data):
+        if "id" in data:
+            self._id = data["id"]
+        if "status" in data:
+            self._status = data["status"]
+        if "started_on" in data:
+            self._started_on = data["started_on"]
+        if "completed_on" in data:
+            self._completed_on = data["completed_on"]
+        if "project_id" in data:
+            self._project_id = data["project_id"]
+        if "experiment_id" in data:
+            self._experiment_id = data["experiment_id"]
+        if "progress" in data:
+            self._progress = data["progress"]
+
+    @staticmethod
+    def _create(requester: Requester, project_id: str, experiment_id: str, confidence_threshold: float = 0.8,
+                max_detections_per_image: int = 100, num_images: int = 0, masker_threshold: float = 0.5,
+                dataset_id: Optional[str] = None):
+        res = requester.post(AutomatedLabelingJob.endpoint.format(project_id=project_id),
+                             json_data={"experiment_id": experiment_id,
+                                        "confidence_threshold": confidence_threshold,
+                                        "max_detections_per_image": max_detections_per_image,
+                                        "num_images": num_images,
+                                        "masker_threshold": masker_threshold,
+                                        "dataset_id": dataset_id})
+        return AutomatedLabelingJob(requester, res, {"project_id": project_id})
+
+    def check_status(self):
+        """
+        Returns current status and progress
+        """
+        if self._last_check is None or self._last_check > WAIT_INTERVAL_SEC:
+            res = self._requester.get(AutomatedLabelingJob.endpoint_job_id.format(project_id=self._project_id,
+                                                                                  job_id=self._id))
+            self._set_prop_values(res)
+            self._last_check = time()
+        return {"status": self._status,
+                "progress": self._progress}

hasty/project.py

@@ -2,6 +2,7 @@
 from typing import List, Union, Optional
 
 from .attribute import Attribute
+from .automated_labeling import AutomatedLabelingJob
 from .constants import ImageStatus, SemanticFormat, VALID_EXPORT_FORMATS, VALID_SEMANTIC_ORDER, \
     VALID_SEMANTIC_FORMATS, VALID_STATUSES
 from .dataset import Dataset
@@ -419,3 +420,39 @@ def get_attributer(self):
         d = Attributer(self._requester, {"project_id": self._id})
         d.discover_model()
         return d
+
+    def get_automated_labeling_jobs(self):
+        """
+        Get automated labeling jobs, list of :py:class:`~hasty.AutomatedLabelingJob` objects.
+        """
+        return PaginatedList(AutomatedLabelingJob, self._requester,
+                             AutomatedLabelingJob.endpoint.format(project_id=self._id),
+                             obj_params={"project_id": self.id})
+
+    def get_automated_labeling_job(self, job_id: str):
+        """
+        Get automated labeling jobs, list of :py:class:`~hasty.AutomatedLabelingJob` objects.
+
+        Args:
+            job_id (str): Automated labeling job id
+        """
+        res = self._requester.get(AutomatedLabelingJob.endpoint_job_id.format(project_id=self.id, job_id=job_id))
+        return AutomatedLabelingJob(self._requester, res, {"project_id": self.id})
+
+    def create_automated_labeling_job(self, experiment_id: str, confidence_threshold: float = 0.8,
+                                      max_detections_per_image: int = 100, num_images: int = 0,
+                                      masker_threshold: float = 0.5, dataset_id: Optional[str] = None):
+        """
+        Create automated labeling job, returns :py:class:`~hasty.AutomatedLabelingJob` object.
+
+        Args:
+            experiment_id (str): ID of an experiment, that would be used for automated labeling
+            confidence_threshold (float, optional): Confidence threshold of the predictions that should be applied
+                                                    (default 0.8)
+            max_detections_per_image (int, optional): Max number of labels that should be created, default 100
+            num_images (int, optional): Total number of images that should be used for automated labeling (default all)
+            masker_threshold (float, optional): Threshold for mask head (default 0.5)
+            dataset_id (str, optional): Filter images by dataset
+        """
+        return AutomatedLabelingJob._create(self._requester, self._id, experiment_id, confidence_threshold,
+                                            max_detections_per_image, num_images, masker_threshold, dataset_id)