Updated TODO

CONFIGURATION.md

@@ -122,6 +122,11 @@ Configuration values in this section are settings specific to real-time mode.
             - `alert` is a decimal number that determines how long Predator will delay before starting the next round when there is an active alert.
             - `normal` is a decimal number that determines how long Predator will delay before starting the next round under normal circumstances.
         - `clearing` is a boolean that determines whether or not Predator will clear the output screen between analysis rounds during real-time mode.
+- `object_recognition` contains settings related to object recognition in real-time mode.
+    - `enabled` is a boolean value that enables and disables object recognition in real-time mode.
+        - This setting does not override the `general>object_recognition>enabled` setting.
+    - `video_still_path` is an absolute file-path to the image Predator should run object recognition on.
+        - By default, Phantom stores video stills from the capture device stream to `/dev/shm/phantom-webcam.jpg`, so that is where this setting should point to in most cases.
 - `gps` contains settings related to GPS-based features.
     - `enabled` is a boolean determines whether GPS features are enabled or disabled.
     - `alpr_location_tagging` is a boolean that determines whether or not the current GPS location will be saved to the log file each time a plate is logged.

TODO.md

@@ -70,5 +70,5 @@ These are the features actively planned for Predator and are likely to be added
 - [X] Improve the efficiency of GPS location requests when many requests are made in quick succession.
     - [ ] Test that improved GPS location querying behaves as expected.
 - [X] Kill the ALPR process every time Predator starts to ensure there are no unexpected background threads.
-- [ ] Re-implement object recogntion to real-time mode using the new back-end.
-- [ ] Re-implement image saving in real-time mode using the new-backend.
+- [X] Re-implement object recogntion to real-time mode using the new back-end.
+    [ ] Verify object recognition functionality.

main.py

@@ -783,7 +783,7 @@
 
 
 
-    # If enabled, count how many vehicles are in each frame.
+    # If enabled, count how many objects are in each frame.
     if (config["general"]["object_recognition"]["enabled"] == True):
         debug_message("Running object recognition")
         print("Running object recognition...")
@@ -1312,6 +1312,27 @@
 
 
 
+        # If enabled, run object recognition on the captured frame.
+        if (config["general"]["object_recognition"]["enabled"] == True and config["realtime"]["object_recognition"]["enabled"] == True): # Check to make sure real-time object recognition is enabled.
+            debug_message("Running object recognition")
+            if (config["realtime"]["interface"]["display"]["output_level"] >= 3): # Only display this status message if the output level indicates to do so.
+                print("Running object recognition...")
+
+            image = cv2.imread(config["realtime"]["object_recognition"]["enabled"]) # Load the frame.
+            object_recognition_bounding_box, object_recognition_labels, object_recognition_confidence = cv.detect_common_objects(image) # Anaylze the image.
+            objects_identified = str(object_recognition_labels) # Convert the list of objects identified into a plain string.
+            if (objects_identified != "[]"): # Check to see that there were actually identified objects.
+                if (config["realtime"]["interface"]["display"]["output_level"] >= 2): # Only display this status message if the output level indicates to do so.
+                    print("Objects identified: " + objects_identified)
+                export_data = str(round(time.time()*10)/10) + "," + objects_identified + "\n" # Add the timestamp to the export data, followed by the object's detected, followed by a line break to prepare for the next entry to be added later.
+                if (save_real_time_object_recognition == True): # Check to make sure the user has configured Predator to save recognized objects to disk.
+                    add_to_file(root + "/real_time_object_detection.csv", export_data, silence_file_saving) # Add the export data to the end of the file and write it to disk.
+
+            if (config["realtime"]["interface"]["display"]["output_level"] >= 3): # Only display this status message if the output level indicates to do so.
+                print("Done\n----------")
+
+
+
 
 
         debug_message("Processing ALPR results")