added docker-compose for local testing; fixed main processing function

.dockerignore

@@ -3,4 +3,8 @@
 /data
 __pycache__
 .coverage
-.flake8
+.flake8
+/config
+/config.yml
+pyproject.toml
+poetry.lock

.gitignore

@@ -1,5 +1,6 @@
 *.mp4
 /config-local.yml
-/data
+/data/*
+!/data/README.md
 __pycache__
 .coverage

config.yml → config/config.yml

@@ -32,7 +32,7 @@ VISXP_PREP:
     RUN_HECATE: true
     RUN_KEYFRAME_EXTRACTION: true
     RUN_AUDIO_EXTRACTION: true
-    TEST_INPUT_FILE: /data/test_openbeelden.mp4  # or null
+    TEST_INPUT_FILE: /data/testob.mp4  # or null
 INPUT:
     DELETE_ON_COMPLETION: True
 OUTPUT:

data/README.md

@@ -0,0 +1 @@
+Put a file in this dir and fill in the path into the config.yml VISXP_PREP.TEST_INPUT_FILE

docker-compose.yml

@@ -0,0 +1,13 @@
+version: '3'
+services:
+  web:
+    image: dane-video-segmentation-worker:latest
+    volumes:
+      - ./data:/data
+      - ./config:/root/.DANE
+    container_name: visxp
+    # command: --cfg=/mnt/data/config.yml
+    logging:
+      options:
+        max-size: 20m
+    restart: always

docker-entrypoint.sh

@@ -2,7 +2,7 @@
 
 # echo "Starting virtual env and DANE video segmentation worker"
 
-python3.10 work_it_locally.py
+python3.10 visxp_prep.py
 # python3.10 worker.py
 
 echo the worker crashed, tailing /dev/null for debugging

visxp_prep.py

@@ -24,7 +24,7 @@ def generate_input_for_feature_extraction(
 
     output_dirs = {}
     for kind in ["keyframes", "metadata", "spectograms", "tmp"]:
-        output_dir = os.path.join("/data", os.path.basename(input_file_path), kind)
+        output_dir = os.path.join("/data", _get_source_id(input_file_path), kind)
         if not os.path.isdir(output_dir):
             os.makedirs(output_dir)
         output_dirs[kind] = output_dir
@@ -39,7 +39,7 @@ def generate_input_for_feature_extraction(
         except Exception:
             logger.info("Could not obtain shots and keyframes. Exit.")
             sys.exit()
-        fps = get_fps(input_file_path)
+        fps = _get_fps(input_file_path)
         logger.info(f"Framerate is {fps}.")
         with open(
             os.path.join(output_dirs["metadata"], "shot_boundaries_timestamps_ms.txt"),
@@ -49,8 +49,8 @@ def generate_input_for_feature_extraction(
                 str(
                     [
                         (
-                            frame_index_to_timecode(start, fps),
-                            frame_index_to_timecode(end, fps),
+                            _frame_index_to_timecode(start, fps),
+                            _frame_index_to_timecode(end, fps),
                         )
                         for (start, end) in shots
                     ]
@@ -63,7 +63,7 @@ def generate_input_for_feature_extraction(
         with open(
             os.path.join(output_dirs["metadata"], "keyframes_timestamps_ms.txt"), "w"
         ) as f:
-            f.write(str([frame_index_to_timecode(i, fps) for i in keyframes]))
+            f.write(str([_frame_index_to_timecode(i, fps) for i in keyframes]))
 
     if cfg.VISXP_PREP.RUN_KEYFRAME_EXTRACTION:
         logger.info("Extracting keyframe images now.")
@@ -106,12 +106,17 @@ def generate_input_for_feature_extraction(
     return VisXPFeatureExtractionInput(500, "Not implemented yet!", -1)
 
 
-def frame_index_to_timecode(frame_index: int, fps: float, out_format="ms"):
+def _get_source_id(input_file_path: str) -> str:
+    fn = os.path.basename(input_file_path)
+    return fn[0 : fn.rfind(".")] if "." in fn else fn
+
+
+def _frame_index_to_timecode(frame_index: int, fps: float, out_format="ms"):
     if out_format == "ms":
         return round(frame_index / fps * 1000)
 
 
-def get_fps(media_file):
+def _get_fps(media_file):
     return cv2.VideoCapture(media_file).get(cv2.CAP_PROP_FPS)