Add some options to start_prep_train_data_job

These options control how the training dataset is generated
from the raw annotations.

Author: lewfish

rasterfoundry/api.py

@@ -8,9 +8,8 @@
 
 from .models import Project, MapToken
 from .exceptions import RefreshTokenException
-from .settings import (
-    RV_CPU_JOB_DEF, RV_CPU_QUEUE, DEVELOP_BRANCH, RV_CONFIG_URI_ROOT)
 from .utils import upload_raster_vision_config
+from .settings import RV_PROJ_CONFIG_DIR_URI
 
 SPEC_PATH = os.path.join(os.path.dirname(os.path.realpath(__file__)),
                          'spec.yml')
@@ -156,12 +155,12 @@ def get_project_configs(self, project_ids, annotation_uris):
 
         return project_configs
 
-                                  output_zip_uri,
-                                  config_uri_root=RV_CONFIG_URI_ROOT,
-                                  job_queue=RV_CPU_QUEUE,
-                                  job_definition=RV_CPU_JOB_DEF,
-                                  branch_name=DEVELOP_BRANCH, attempts=1):
     def start_prep_train_data_job(self, rv_batch_client, project_ids,
+                                  annotation_uris,
+                                  output_zip_uri, label_map_uri,
+                                  proj_config_dir_uri=RV_PROJ_CONFIG_DIR_URI,
+                                  min_area=None, single_label=None,
+                                  no_partial=True, channel_order=None):
         """Start a Batch job to prepare object detection training data.
 
         Args:
@@ -170,24 +169,41 @@ def start_prep_train_data_job(self, rv_batch_client, project_ids,
             project_ids (list of str): ids of projects to make train data for
             annotation_uris (list of str): annotation URIs for projects
             output_zip_uri (str): URI of output zip file
-            config_uri_root (str): The root of generated URIs for config files
-            job_queue (str): name of the Batch job queue to run the job in
-            job_definition (str): name of the Batch job definition
-            branch_name (str): branch of the raster-vision repo to use
-            attempts (int): number of attempts for the Batch job
+            label_map_uri (str): URI of output label map
+            proj_config_dir_uri (str): The root of generated URIs for config
+                files
+            min_area (float): minimum area of bounding boxes to include
+            single_label (str): Convert all labels to this label
+            no_partial (bool): Black out partially visible objects
+            channel_order: list of length 3 with GeoTIFF channel indices to
+                map to RGB.
 
         Returns:
             job_id (str): job_id of job started on Batch
         """
         project_configs = self.get_project_configs(
             project_ids, annotation_uris)
         config_uri = upload_raster_vision_config(
-            project_configs, config_uri_root)
-
-        command = ('python -m rv.run prep_train_data --debug ' +
-                   '--chip-size 300 --num-neg-chips 100 ' +
-                   '--max-attempts 500 {} {}').format(
-                       config_uri, output_zip_uri)
+            project_configs, proj_config_dir_uri)
+
+        base_command = \
+            'python -m rv.run prep_train_data --debug --chip-size 300 '
+        min_area_opt = ('--min-area {} '.format(min_area)
+                        if min_area is not None else '')
+        single_label_opt = ('--single-label {} '.format(single_label)
+                            if single_label is not None else '')
+        no_partial_opt = '--no-partial ' if no_partial else ''
+
+        channel_order_opt = ''
+        if channel_order is not None:
+            channel_order_str = ' '.join([
+                str(channel_ind) for channel_ind in channel_order])
+            channel_order_opt = ('--channel-order {} '
+                                 .format(channel_order_str))
+
+        command = (base_command + min_area_opt + single_label_opt +
+                   no_partial_opt + channel_order_opt + '{} {} {}')
+        command = command.format(config_uri, output_zip_uri, label_map_uri)
 
         job_name = 'prep_train_data_{}'.format(uuid.uuid1())
         job_id = rv_batch_client.start_raster_vision_job(job_name, command)