Significantly improve video resolution/quality using ESPCN_x4 model

README.md

@@ -146,6 +146,7 @@ You can now use the virtual camera output (uses pyvirtualcam) by turning on the
 
 Additional command line arguments are given below. To learn out what they do, check [this guide](https://github.com/s0md3v/roop/wiki/Advanced-Options).
 
+
 ```
 options:
   -h, --help                                               show this help message and exit
@@ -167,11 +168,17 @@ options:
   --headless                                               run in headless mode
   --enhancer-upscale-factor                                Sets the upscale factor for the enhancer. Only applies if `face_enhancer` is set as a frame-processor
   --source-image-scaling-factor                            Set the upscale factor for source images. Only applies if `face_swapper` is set as a frame-processor
+  -r SCALE, --super-resolution-scale-factor SCALE          Super resolution scale factor, choices are 2, 3, 4
   -v, --version                                            show program's version number and exit
 ```
 
 Looking for a CLI mode? Using the -s/--source argument will make the run program in cli mode.
 
+To improve the video quality, you can use the `super_resolution` frame processor after swapping the faces. It will enhance the video quality by 2x, 3x or 4x. You can set the upscale factor using the `-r` or `--super-resolution-scale-factor` argument. 
+Processing time will increase with the upscale factor, it's quite quick though.
+
+```
+
 ## Credits
 - [henryruhs](https://github.com/henryruhs): for being an irreplaceable contributor to the project
 - [ffmpeg](https://ffmpeg.org/): for making video related operations easy

modules/core.py

@@ -51,14 +51,19 @@ def parse_args() -> None:
     program.add_argument('-t', '--target', help='Select a target image or video', dest='target_path')
     program.add_argument('-o', '--output', help='Select output file or directory', dest='output_path')
     program.add_argument('--frame-processor', help='Pipeline of frame processors', dest='frame_processor',
-                         default=['face_swapper'], choices=['face_swapper', 'face_enhancer', 'super_resolution'], nargs='+')
+                         default=['face_swapper'], choices=['face_swapper', 'face_enhancer', 'super_resolution'],
+                         nargs='+')
     program.add_argument('--keep-fps', help='Keep original fps', dest='keep_fps', action='store_true', default=False)
-    program.add_argument('--keep-audio', help='Keep original audio', dest='keep_audio', action='store_true', default=True)
+    program.add_argument('--keep-audio', help='Keep original audio', dest='keep_audio', action='store_true',
-    program.add_argument('--keep-frames', help='Keep temporary frames', dest='keep_frames', action='store_true', default=False)
+                         default=True)
-    program.add_argument('--many-faces', help='Process every face', dest='many_faces', action='store_true', default=False)
+    program.add_argument('--keep-frames', help='Keep temporary frames', dest='keep_frames', action='store_true',
+                         default=False)
+    program.add_argument('--many-faces', help='Process every face', dest='many_faces', action='store_true',
+                         default=False)
     program.add_argument('--video-encoder', help='Adjust output video encoder', dest='video_encoder', default='libx264',
                          choices=['libx264', 'libx265', 'libvpx-vp9'])
-    program.add_argument('--video-quality', help='Adjust output video quality', dest='video_quality', type=int, default=18,
+    program.add_argument('--video-quality', help='Adjust output video quality', dest='video_quality', type=int,
+                         default=18,
                          choices=range(52), metavar='[0-51]')
     program.add_argument('--live-mirror', help='The live camera display as you see it in the front-facing camera frame',
                          dest='live_mirror', action='store_true', default=False)
@@ -76,6 +81,8 @@ def parse_args() -> None:
                          dest='enhancer_upscale_factor', type=int, default=1)
     program.add_argument('--source-image-scaling-factor', help='Set the upscale factor for source images',
                          dest='source_image_scaling_factor', default=2, type=int)
+    program.add_argument('-r', '--super-resolution-scale-factor', dest='super_resolution_scale_factor',
+                         help='Set the upscale factor for super resolution', default=4, choices=[2, 3, 4], type=int)
     program.add_argument('-v', '--version', action='version',
                          version=f'{modules.metadata.name} {modules.metadata.version}')
 
@@ -107,6 +114,7 @@ def parse_args() -> None:
     modules.globals.headless = args.headless
     modules.globals.enhancer_upscale_factor = args.enhancer_upscale_factor
     modules.globals.source_image_scaling_factor = args.source_image_scaling_factor
+    modules.globals.sr_scale_factor = args.super_resolution_scale_factor
     # Handle face enhancer tumbler
     modules.globals.fp_ui['face_enhancer'] = 'face_enhancer' in args.frame_processor
 
@@ -191,12 +199,14 @@ def limit_resources() -> None:
             try:
                 soft, hard = resource.getrlimit(resource.RLIMIT_DATA)
                 if memory > hard:
-                    print(f"Warning: Requested memory limit {memory / (1024 ** 3)} GB exceeds system's hard limit. Setting to maximum allowed {hard / (1024 ** 3)} GB.")
+                    print(
+                        f"Warning: Requested memory limit {memory / (1024 ** 3)} GB exceeds system's hard limit. Setting to maximum allowed {hard / (1024 ** 3)} GB.")
                     memory = hard
                 resource.setrlimit(resource.RLIMIT_DATA, (memory, memory))
             except ValueError as e:
                 print(f"Warning: Could not set memory limit: {e}. Continuing with default limits.")
 
+
 def release_resources() -> None:
     if 'cuda' in modules.globals.execution_providers:
         torch.cuda.empty_cache()
@@ -250,7 +260,8 @@ def process_image_to_image() -> None:
 
     for frame_processor in get_frame_processors_modules(modules.globals.frame_processors):
         update_status('Processing...', frame_processor.NAME)
-        frame_processor.process_image(modules.globals.source_path, modules.globals.output_path, modules.globals.output_path)
+        frame_processor.process_image(modules.globals.source_path, modules.globals.output_path,
+                                      modules.globals.output_path)
         release_resources()
 
     if is_image(modules.globals.target_path):

modules/globals.py

@@ -31,4 +31,5 @@ nsfw = None
 camera_input_combobox = None
 webcam_preview_running = False
 enhancer_upscale_factor = 1
-source_image_scaling_factor = 2
+source_image_scaling_factor = 2
+sr_scale_factor = 4

modules/processors/frame/super_resolution.py

@@ -20,7 +20,7 @@ class SuperResolutionModel:
     _instance = None
     _lock = threading.Lock()
 
-    def __init__(self, sr_model_path: str = 'ESPCN_x4.pb'):
+    def __init__(self, sr_model_path: str = f'ESPCN_x{modules.globals.sr_scale_factor}.pb'):
         if SuperResolutionModel._instance is not None:
             raise Exception("This class is a singleton!")
         self.sr = cv2.dnn_superres.DnnSuperResImpl_create()
@@ -29,21 +29,20 @@ class SuperResolutionModel:
             raise FileNotFoundError(f"Super-resolution model not found at {self.model_path}")
         try:
             self.sr.readModel(self.model_path)
-            self.sr.setModel("espcn", 4)  # Using ESPCN with 4x upscaling
+            self.sr.setModel("espcn", modules.globals.sr_scale_factor)  # Using ESPCN with 2,3 or 4x upscaling
         except Exception as e:
             print(f"Error during super-resolution model initialization: {e}")
             raise e
 
     @classmethod
-    def get_instance(cls, sr_model_path: str = 'ESPCN_x4.pb'):
+    def get_instance(cls, sr_model_path: str = f'ESPCN_x{modules.globals.sr_scale_factor}.pb'):
         if cls._instance is None:
             with cls._lock:
                 if cls._instance is None:
                     try:
                         cls._instance = cls(sr_model_path)
                     except Exception as e:
-                        print(f"Failed to initialize SuperResolutionModel: {e}")
+                        raise RuntimeError(f"Failed to initialize SuperResolution: {str(e)}")
-                        return None
         return cls._instance
 
 
@@ -54,7 +53,7 @@ def pre_check() -> bool:
     download_directory_path = resolve_relative_path('../models')
     # Download the super-resolution model as well
     conditional_download(download_directory_path, [
-        'https://huggingface.co/spaces/PabloGabrielSch/AI_Resolution_Upscaler_And_Resizer/resolve/bcd13b766a9499196e8becbe453c4a848673b3b6/models/ESPCN_x4.pb'
+        f'https://huggingface.co/spaces/PabloGabrielSch/AI_Resolution_Upscaler_And_Resizer/resolve/bcd13b766a9499196e8becbe453c4a848673b3b6/models/ESPCN_x{modules.globals.sr_scale_factor}.pb'
     ])
     return True