Implement caching of the GPS data to reduce latency

It appears that the GPS can take up to 1 second to update results.
This is way too much latency when trying to capture fast-moving objects like birds.

Author: jwillikers

README.adoc

@@ -297,6 +297,7 @@ just --list
 Can't cache the time and date from the GPS then.
 Use TTLCache from https://github.com/tkem/cachetools/[cachetools].
 - [] Add support for a TOML config file with https://github.com/bw2/ConfigArgParse[ConfigArgParse].
+- [] async
 - [] Switch from picamera2 to gstreamer to work with more hardware.
 - [] mypy
 - [] Create a weatherproof enclosure for the camera.

detectionator.py

@@ -1,5 +1,6 @@
 #!/usr/bin/env python3
 import argparse
+from cachetools import cached, TTLCache
 from dateutil import parser
 from functools import partial
 import logging
@@ -114,6 +115,9 @@ def scale(coord, scaler_crop_maximum, lores):
 
 
 # Retrieve and format the data from the GPS for EXIF.
+# Only update the GPS data every 5 minutes to reduce latency.
+# Retrieving data from the GPS can take up to one second, which is too long.
+@cached(cache=TTLCache(maxsize=1, ttl=300))
 def get_gps_exif_metadata(session: gps.gps) -> dict:
     while True:
         if session.read() != 0:
@@ -235,17 +239,17 @@ def main():
     )
     args = parser.parse_args()
 
+    numeric_log_level = getattr(logging, args.log_level.upper(), None)
+    if not isinstance(numeric_log_level, int):
+        raise ValueError(f"Invalid log level: {args.log_level}")
+    logging.basicConfig(level=numeric_log_level)
+
     if args.burst < 1:
         logging.warn(
             f"The burst value must be at least 1. Ignoring the provided burst value of '{args.burst}'."
         )
         args.burst = 1
 
-    numeric_log_level = getattr(logging, args.log_level.upper(), None)
-    if not isinstance(numeric_log_level, int):
-        raise ValueError(f"Invalid log level: {args.log_level}")
-    logging.basicConfig(level=numeric_log_level)
-
     output_directory = os.path.join(os.getenv("HOME"), "Pictures")
     if args.output:
         output_directory = args.output
@@ -397,6 +401,9 @@ def capture_sample_signal_handler(_sig, _frame):
             image = picam2.capture_array("lores")
             matches = inference_tensorflow(image, args.model, labels, match)
             if len(matches) == 0:
+                # Retrieve the GPS data to ensure the cache is up-to-date in order to reduce latency when there is a detection.
+                # todo Update the GPS data asynchronously to allow the detection process to continue uninterrupted instead of blocking when there is a cache miss.
+                get_gps_exif_metadata(gps_session)
                 # Take a quick breather to give the CPU a break.
                 # todo Increase / decrease this wait based on recent detections.
                 time.sleep(0.2)

requirements-dev.in

@@ -3,4 +3,5 @@ pip-tools
 pre-commit
 pytest
 ruff
+types-cachetools
 yamllint

requirements-dev.txt

@@ -163,6 +163,11 @@ ruff==0.4.3 \
     --hash=sha256:eeb039f8428fcb6725bb63cbae92ad67b0559e68b5d80f840f11914afd8ddf7f \
     --hash=sha256:ff0a3ef2e3c4b6d133fbedcf9586abfbe38d076041f2dc18ffb2c7e0485d5a07
     # via -r requirements-dev.in
+types-cachetools==5.3.0.7 \
+    --hash=sha256:27c982cdb9cf3fead8b0089ee6b895715ecc99dac90ec29e2cab56eb1aaf4199 \
+    --hash=sha256:2c6d19900131b36d4fd5d7b316fd060870145153ead45e29dafbe10538533008 \
+    --hash=sha256:98c069dc7fc087b1b061703369c80751b0a0fc561f6fb072b554e5eee23773a0
+    # via -r requirements-dev.in
 virtualenv==20.26.1 \
     --hash=sha256:604bfdceaeece392802e6ae48e69cec49168b9c5f4a44e483963f9242eb0e78b \
     --hash=sha256:71c11a0f2ec69d818248974a7f948ee69655547ddc2148f45dd204e73eaaed04 \

requirements.in

@@ -1,3 +1,4 @@
+cachetools
 numpy
 opencv-python
 piexif