DockerFile for Mac users to run DeepfaceLab with CPU Mode (#95)

* fix localization nullpointer exception

* fix devicelib error line:61,remove e

* support create docker from cpu dockerfile

* support preview or not when train(resolve cannot connect to X server)

.gitignore

@@ -12,4 +12,6 @@
 !mathlib
 !models
 !nnlib
-!utils
+!utils
+!Dockerfile*
+!*.sh

DockerCPU.md

@@ -0,0 +1,132 @@
+# For Mac Users
+If you just have a **MacBook**.DeepFaceLab **GPU** mode does not works. However,it can also works with **CPU** mode.Follow the Steps below will help you build the **DRE** (DeepFaceLab Runtime Environment) Easier.
+
+### 1. Open a new terminal and Clone DeepFaceLab with git
+```
+$ git git@github.com:iperov/DeepFaceLab.git
+```
+
+### 2. Change the directory to DeepFaceLab
+```
+$ cd DeepFaceLab
+```
+
+### 3. Install Docker
+
+[Docker Desktop for Mac](https://hub.docker.com/editions/community/docker-ce-desktop-mac)
+
+### 4. Build Docker Image For DeepFaceLab
+
+```
+$ docker build -t deepfacelab-cpu -f Dockerfile.cpu .
+```
+
+### 5. Mount DeepFaceLab volume and Run it
+
+```
+$ docker run -p 8888:8888  --hostname deepfacelab-cpu --name deepfacelab-cpu  -v $PWD:/notebooks  deepfacelab-cpu
+```
+
+PS: Because your current directory is `DeepFaceLab`,so `-v $PWD:/notebooks` means Mount `DeepFaceLab` volume to `notebooks` in **Docker**
+
+And then you will see the log below:
+
+```
+The Jupyter Notebook is running at:
+http://(deepfacelab-cpu or 127.0.0.1):8888/?token=your token
+```
+
+### 6. Open a new terminal to run DeepFaceLab in /notebooks
+
+```
+$ docker exec -it deepfacelab-cpu bash
+$ ls -A
+```
+
+### 7. Use jupyter in deepfacelab-cpu bash
+
+```
+$ jupyter notebook list
+```
+or just open it on your browser `http://127.0.0.1:8888/?token=your_token`
+
+PS: You can run python with jupyter.However,we just run our code in bash.It's simpler and clearer.Now the **DRE** (DeepFaceLab Runtime Environment) almost builded.
+
+### 8. Stop or Kill Docker Container
+
+```
+$ docker stop deepfacelab-cpu
+$ docker kill deepfacelab-cpu
+```
+
+### 9. Start Docker Container
+
+```
+# start docker container
+$ docker start deepfacelab-cpu
+# open bash to run deepfacelab
+$ docker exec -it deepfacelab-cpu bash
+```
+
+PS: `STEP 8` or `STEP 9` just show you the way to stop and start **DRE**.
+
+### 10. enjoy it
+
+```
+# make sure you current directory is `/notebooks`
+$ pwd
+# make sure all `DeepFaceLab` code is in current path `/notebooks`
+$ ls -a
+# read and write permission
+$ chmod +x cpu.sh
+# run `DeepFaceLab`
+$ ./cpu.sh
+```
+
+### Details with `DeepFaceLab`
+
+#### 1. Concepts
+
+![SRC](doc/DF_Cage_0.jpg)
+
+In our Case,**Cage**'s Face is **SRC Face**,and **Trump**'s Face is **DST Face**.and finally we get the **Result** below.
+
+![Result](doc/merged-face.jpg)
+
+So,before you run `./cpu.sh`.You should be aware of this.
+
+#### 2. Use MTCNN(mt) to extract faces
+Do not use DLIB extractor in CPU mode
+
+#### 3. Best practice for SORT
+1) delete first unsorted aligned groups of images what you can to delete.
+
+2) use `hist`
+
+#### 4. Use `H64 model` to train and convert
+Only H64 model reasonable to train on home CPU.You can choice other  model like **H128 (3GB+)** | **DF (5GB+)** and so on ,it depends entirely on your CPU performance.
+
+#### 5. execute the script below one by one
+
+```
+root@deepfacelab-cpu:/notebooks# ./cpu.sh
+1) clear workspace		      7) data_dst sort by hist
+2) extract PNG from video data_src    8) train
+3) data_src extract faces	      9) convert
+4) data_src sort		     10) converted to mp4
+5) extract PNG from video data_dst   11) quit
+6) data_dst extract faces
+Please enter your choice:       
+```
+
+#### 6. Put all videos in `workspace` directory
+```
+.
+├── data_dst
+├── data_src
+├── dst.mp4
+├── model
+└── src.mp4
+
+3 directories, 2 files
+```

Dockerfile.cpu

@@ -0,0 +1,17 @@
+FROM tensorflow/tensorflow:latest-py3
+
+RUN apt-get update -qq -y \
+ && apt-get install -y libsm6 libxrender1 libxext-dev python3-tk\
+ && apt-get install -y ffmpeg \
+ && apt-get install -y wget \
+ && apt-get install -y vim \
+ && apt-get install -y git \
+ && apt-get clean \
+ && rm -rf /var/lib/apt/lists/*
+
+COPY requirements-cpu-docker.txt /opt/
+RUN pip3 install cmake
+RUN pip3 --no-cache-dir install -r /opt/requirements-cpu-docker.txt && rm /opt/requirements-cpu-docker.txt
+
+WORKDIR "/notebooks"
+CMD ["/run_jupyter.sh", "--allow-root"]

README.md

@@ -176,6 +176,9 @@ Video tutorial: https://www.youtube.com/watch?v=K98nTNjXkq8
 
 Windows 10 consumes % of VRAM even if card unused for video output.
 
+### For Mac Users
+Check out [DockerCPU.md](DockerCPU.md) for more detailed instructions.
+
 ### **Problem of the year**:
 
 algorithm of overlaying neural face onto video face located in ConverterMasked.py.

cpu.sh

@@ -0,0 +1,71 @@
+#!/bin/bash
+INTERNAL_DIR=`pwd`
+WORKSPACE=$INTERNAL_DIR/workspace
+PYTHON=`which python`
+
+PS3="Please enter your choice: "
+options=("clear workspace" "extract PNG from video data_src" "data_src extract faces" "data_src sort" "extract PNG from video data_dst" "data_dst extract faces" "data_dst sort by hist" "train" "convert" "converted to mp4" "quit")
+select opt in "${options[@]}"
+do
+        case $opt in
+                "clear workspace" )
+                        echo -n "Clean up workspace? [Y/n] "; read workspace_ans
+                        if [ "$workspace_ans" == "Y" ] || [ "$workspace_ans" == "y" ]; then
+                                rm -rf $WORKSPACE
+                                mkdir -p $WORKSPACE/data_src/aligned
+                                mkdir -p $WORKSPACE/data_dst/aligned
+                                mkdir -p $WORKSPACE/model
+                                echo "Workspace has been successfully cleaned!"
+                        fi
+                        ;;
+                "extract PNG from video data_src" )
+                        echo -n "File name: "; read filename
+                        echo -n "FPS: "; read fps
+                        if [ -z "$fps" ]; then fps="25"; fi
+                        ffmpeg -i $WORKSPACE/$filename -r $fps $WORKSPACE/data_src/%04d.png -loglevel error
+                        ;;
+                "data_src extract faces" )
+                        echo -n "Detector? [mt | manual] "; read detector
+                        $PYTHON $INTERNAL_DIR/main.py extract --input-dir $WORKSPACE/data_src --output-dir $WORKSPACE/data_src/aligned --detector $detector --debug --cpu-only
+                        ;;
+                "data_src sort" )
+                        echo -n "Sort by? [blur | brightness | face-yaw | hue | hist | hist-blur | hist-dissim] "; read sort_method
+                        $PYTHON $INTERNAL_DIR/main.py sort --input-dir $WORKSPACE/data_src/aligned --by $sort_method
+                        ;;
+                "extract PNG from video data_dst" )
+                        echo -n "File name: "; read filename
+                        echo -n "FPS: "; read fps
+                        if [ -z "$fps" ]; then fps="25"; fi
+                        ffmpeg -i $WORKSPACE/$filename -r $fps $WORKSPACE/data_dst/%04d.png -loglevel error
+                        ;;
+                "data_dst extract faces" )
+                        echo -n "Detector? [mt | manual] "; read detector
+                        $PYTHON $INTERNAL_DIR/main.py extract --input-dir $WORKSPACE/data_dst --output-dir $WORKSPACE/data_dst/aligned --detector $detector --debug --cpu-only
+                        ;;
+                "data_dst sort by hist" )
+                        $PYTHON $INTERNAL_DIR/main.py sort --input-dir $WORKSPACE/data_dst/aligned --by hist
+                        ;;
+                "train" )
+                        echo -n "Model? [ H64 (2GB+) | H128 (3GB+) | DF (5GB+) | LIAEF128 (5GB+) | LIAEF128YAW (5GB+) | MIAEF128 (5GB+) | AVATAR (4GB+) ] "; read model
+                        echo -n "Show Preview? [Y/n] "; read preview
+                        if [ "$preview" == "Y" ] || [ "$preview" == "y" ]; then preview="--preview"; else preview=""; fi
+                        $PYTHON $INTERNAL_DIR/main.py train --training-data-src-dir $WORKSPACE/data_src/aligned --training-data-dst-dir $WORKSPACE/data_dst/aligned --model-dir $WORKSPACE/model --model $model --cpu-only $preview
+                        ;;
+                "convert" )
+                        echo -n "Model? [ H64 (2GB+) | H128 (3GB+) | DF (5GB+) | LIAEF128 (5GB+) | LIAEF128YAW (5GB+) | MIAEF128 (5GB+) | AVATAR(4GB+) ] "; read model
+                        $PYTHON $INTERNAL_DIR/main.py convert --input-dir $WORKSPACE/data_dst --output-dir $WORKSPACE/data_dst/merged --aligned-dir $WORKSPACE/data_dst/aligned --model-dir $WORKSPACE/model --model $model --ask-for-params --cpu-only
+                        ;;
+                "converted to mp4" )
+                        echo -n "File name of destination video: "; read filename
+                        echo -n "FPS: "; read fps
+                        if [ -z "$fps" ]; then fps="25"; fi
+                        ffmpeg -y -i $WORKSPACE/$filename -r $fps -i "$WORKSPACE/data_dst/merged/%04d.png" -map 0:a? -map 1:v -r $fps -c:v libx264 -b:v 8M -pix_fmt yuv420p -c:a aac -strict -2 -b:a 192k -ar 48000 "$WORKSPACE/result.mp4" -loglevel error
+                        ;;
+                "quit" )
+                        break
+                        ;;
+                *)
+                        echo "Invalid choice!"
+                        ;;
+        esac
+done

localization/localization.py

@@ -2,7 +2,8 @@
 import locale
 
 system_locale = locale.getdefaultlocale()[0]
-system_language = system_locale[0:2]
+# system_locale may be nil
+system_language = system_locale[0:2] if system_locale is not None else "en"
 
 windows_font_name_map = {
     'en' : 'cour',

main.py

@@ -80,6 +80,7 @@ if __name__ == "__main__":
             model_path=arguments.model_dir, 
             model_name=arguments.model_name,
             debug              = arguments.debug,
+            preview           = arguments.preview,
             #**options
             batch_size         = arguments.batch_size,
             write_preview_history = arguments.write_preview_history,
@@ -106,8 +107,9 @@ if __name__ == "__main__":
     train_parser.add_argument('--force-best-gpu-idx', type=int, dest="force_best_gpu_idx", default=-1, help="Force to choose this GPU idx as best(worst).")
     train_parser.add_argument('--multi-gpu', action="store_true", dest="multi_gpu", default=False, help="MultiGPU option. It will select only same best(worst) GPU models.")
     train_parser.add_argument('--force-gpu-idxs', type=str, dest="force_gpu_idxs", default=None, help="Override final GPU idxs. Example: 0,1,2.")
-    train_parser.add_argument('--cpu-only', action="store_true", dest="cpu_only", default=False, help="Train on CPU.")    
+    train_parser.add_argument('--cpu-only', action="store_true", dest="cpu_only", default=False, help="Train on CPU.")
-    
+    train_parser.add_argument('--preview', action="store_true",dest="preview", default=False, help="Show preview.")
+
     train_parser.set_defaults (func=process_train)
     
     def process_convert(arguments):

mainscripts/Trainer.py

@@ -277,13 +277,14 @@ def previewThread (input_queue, output_queue):
             
     cv2.destroyAllWindows()
     
-def main (training_data_src_dir, training_data_dst_dir, model_path, model_name, **in_options):
+def main (training_data_src_dir, training_data_dst_dir, model_path, model_name,preview, **in_options):
-    print ("Running trainer.\r\n")
+    print ("Running trainer(preview=%s).\r\n" % (preview))
     
     output_queue = queue.Queue()
     input_queue = queue.Queue()
     import threading
     thread = threading.Thread(target=trainerThread, args=(output_queue, input_queue, training_data_src_dir, training_data_dst_dir, model_path, model_name), kwargs=in_options )
     thread.start()
-    
+
-    previewThread (input_queue, output_queue)
+    if preview:
+        previewThread (input_queue, output_queue)

nnlib/devicelib.py

@@ -58,7 +58,7 @@ class devicelib:
         try:
             nvmlInit()
             nvmlShutdown()
-        except e:
+        except:
             return False
         return True    
      

requirements-cpu-docker.txt

@@ -0,0 +1,9 @@
+pathlib==1.0.1
+scandir==1.6
+h5py==2.7.1
+Keras==2.2.4
+opencv-python==3.4.0.12
+scikit-image
+dlib==19.10.0
+tqdm
+git+https://www.github.com/keras-team/keras-contrib.git