Make quality adjustable

- switch to imageio for video writing
- make error messages more expressive

README.md

@@ -57,7 +57,6 @@ To get a local copy up and running follow these simple steps.
 
 You need a working Python environment with a Python version of 3.8 or higher that satisfies the listed `requirements.txt`. Depending on your machine, you can leverage GPU acceleration for pytorch - see [here](https://pytorch.org/get-started/locally/) or just use `requirements-gpu.txt`.
 
-Further, h264 was chosen for the output video file. You can download an open source library to add support [here](https://github.com/cisco/openh264/releases). OpenCV currently expects version 1.8.0.
 
 Since OpenCV does not care about audio channels, ffmpeg is used to combine the edited video and the audio channel of the input video. The environment variable `FFMPEG_BINARY` needs to be set to the ffmpeg executable for this to work.
 
@@ -73,9 +72,7 @@ Since OpenCV does not care about audio channels, ffmpeg is used to combine the e
    conda activate py38
    pip install -r requirements.txt
    ```
-
-3. Place `openh264-1.8.0-win64.dll` in the same folder as `main.py`.
-
+3. Install ffmpeg binaries (release essentials is enough) and create an environment variable "FFMPEG_BINARY" that points to the ffmpeg.exe binary.
 <!-- USAGE EXAMPLES -->
 ## Usage
 On first launch, the YOLOv5 model is automatically downloaded and fused with the custom weights for face and plate detection from this repo.

dashcamcleaner/main.py

@@ -47,10 +47,20 @@ def setup_blurrer(self):
         self.blurrer.setMaximum.connect(self.setMaximumValue)
         self.blurrer.updateProgress.connect(self.setProgress)
         self.blurrer.finished.connect(self.blurrer_finished)
+        self.blurrer.alert.connect(self.blurrer_alert)
         msg_box = QMessageBox()
         msg_box.setText(f"Successfully loaded {weights_name}.pt")
         msg_box.exec_()
 
+    def blurrer_alert(self, message: str):
+        """
+        Display blurrer messages in the GUI
+        :param message: Message to be displayed
+        """
+        msg_box = QMessageBox()
+        msg_box.setText(message)
+        msg_box.exec_()
+
     def button_abort_clicked(self):
         """
         Callback for button_abort
@@ -84,7 +94,7 @@ def button_start_clicked(self):
         self.ui.button_start.setEnabled(False)
 
         # read inference size
-        inference_size = int(self.ui.combo_box_scale.currentText()[:-1]) * 16 / 9 # ouch again
+        inference_size = int(self.ui.combo_box_scale.currentText()[:-1]) * 16 / 9  # ouch again
 
         # set up parameters
         parameters = {
@@ -94,7 +104,8 @@ def button_start_clicked(self):
             "blur_memory": self.ui.spin_memory.value(),
             "threshold": self.ui.double_spin_threshold.value(),
             "roi_multi": self.ui.double_spin_roimulti.value(),
-            "inference_size": inference_size
+            "inference_size": inference_size,
+            "quality": self.ui.spin_quality.value()
         }
         if self.blurrer:
             self.blurrer.parameters = parameters

dashcamcleaner/src/blurrer.py

@@ -3,6 +3,7 @@
 from timeit import default_timer as timer
 
 import cv2
+import imageio
 import numpy as np
 import torch
 from PySide2.QtCore import QThread, Signal
@@ -12,6 +13,7 @@
 class VideoBlurrer(QThread):
     setMaximum = Signal(int)
     updateProgress = Signal(int)
+    alert = Signal(str)
 
     def __init__(self, weights_name, parameters=None):
         """
@@ -111,6 +113,7 @@ def run(self):
         temp_output = f"{os.path.splitext(self.parameters['output_path'])[0]}_copy{os.path.splitext(self.parameters['output_path'])[1]}"
         output_path = self.parameters["output_path"]
         threshold = self.parameters["threshold"]
+        quality = self.parameters["quality"]
 
         # customize detector
         self.detector.conf = threshold
@@ -125,26 +128,25 @@ def run(self):
         fps = cap.get(cv2.CAP_PROP_FPS)
 
         # save the video to a file
-        fourcc = cv2.VideoWriter_fourcc(*'H264')
-        writer = cv2.VideoWriter(temp_output, fourcc, fps, (width, height))
+        writer = imageio.get_writer(temp_output, codec="libx264", fps=fps, quality=quality)
 
         # update GUI's progress bar on its maximum frames
         self.setMaximum.emit(length)
 
-        if cap.isOpened() == False:
-            print('error file not found')
+        if not cap.isOpened():
+            self.alert.emit('Error: Video file could not be found')
             return
 
         # loop through video
         current_frame = 0
         while cap.isOpened():
             ret, frame = cap.read()
 
-            if ret == True:
+            if ret:
                 new_detections = self.detect_identifiable_information(frame.copy())
                 frame = self.apply_blur(frame, new_detections)
-                writer.write(frame)
-
+                frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
+                writer.append_data(frame_rgb)
             else:
                 break
 
@@ -153,18 +155,25 @@ def run(self):
 
         self.detections = []
         cap.release()
-        writer.release()
+        writer.close()
 
         # copy over audio stream from original video to edited video
         ffmpeg_exe = os.getenv("FFMPEG_BINARY")
+        if not ffmpeg_exe:
+            self.alert.emit(
+                "FFMPEG could not be found! Please make sure the ffmpeg.exe is available under the envirnment variable 'FFMPEG_BINARY'.")
+            return
         subprocess.run(
             [ffmpeg_exe, "-y", "-i", temp_output, "-i", input_path, "-c", "copy", "-map", "0:0", "-map", "1:1",
              "-shortest", output_path])
 
         # delete temporary output that had no audio track
-        os.remove(temp_output)
+        try:
+            os.remove(temp_output)
+        except:
+            self.alert.emit("Could not delete temporary, muted video. Maybe another process (like a cloud service or antivirus) is using it already.")
 
-        ## store sucess and elapsed time
+        # store success and elapsed time
         self.result["success"] = True
         self.result["elapsed_time"] = timer() - start