Merge branch 'improvement/RED-10018' into 'master'

feat: parameterize image stiching tolerance

Closes RED-10018

See merge request redactmanager/image-classification-service!16

config/settings.toml

@@ -5,6 +5,7 @@ level = "INFO"
 # Print document processing progress to stdout
 verbose = false
 batch_size = 16
+image_stiching_tolerance = 1  # in pixels
 mlflow_run_id = "fabfb1f192c745369b88cab34471aba7"
 
 # These variables control filters that are applied to either images, image metadata or service_estimator predictions.

scripts/run_pipeline.py

@@ -36,7 +36,7 @@ def process_pdf(pipeline, pdf_path, page_range=None):
 
 
 def main(args):
-    pipeline = load_pipeline(verbose=True, batch_size=CONFIG.service.batch_size)
+    pipeline = load_pipeline(verbose=CONFIG.service.verbose, batch_size=CONFIG.service.batch_size, tolerance=CONFIG.service.image_stiching_tolerance)
 
     if os.path.isfile(args.input):
         pdf_paths = [args.input]

src/image_prediction/pipeline.py

@@ -3,6 +3,7 @@ from functools import lru_cache, partial
 from itertools import chain, tee
 
 from funcy import rcompose, first, compose, second, chunks, identity, rpartial
+from kn_utils.logging import logger
 from tqdm import tqdm
 
 from image_prediction.config import CONFIG
@@ -21,6 +22,7 @@ os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"
 
 @lru_cache(maxsize=None)
 def load_pipeline(**kwargs):
+    logger.info(f"Loading pipeline with kwargs: {kwargs}")
     model_loader = get_mlflow_model_loader(MLRUNS_DIR)
     model_identifier = CONFIG.service.mlflow_run_id
 

src/serve.py

@@ -18,7 +18,7 @@ logger.reconfigure(sink=stdout, level=CONFIG.logging.level)
 # FIXME: Find more fine-grained solution or if the problem occurs persistently for python services,
 @wrap_in_process
 def process_data(data: bytes, _message: dict) -> list:
-    pipeline = load_pipeline(verbose=CONFIG.service.verbose, batch_size=CONFIG.service.batch_size)
+    pipeline = load_pipeline(verbose=CONFIG.service.verbose, batch_size=CONFIG.service.batch_size, tolerance=CONFIG.service.image_stiching_tolerance)
     return list(pipeline(data))