Updated documentation and requirements

classification/.gitignore

@@ -0,0 +1,4 @@
+env
+classification/network/__pycache__
+classification/dataset/__pycache__
+model/*

classification/README.md

@@ -5,15 +5,38 @@ The models were trained using the Face2Face face tracker, though the `detect_fro
 
 Note that we provide the trained models from our paper which have not been fine-tuned for general compressed videos. You can find our used models under [this link](http://kaldir.vc.in.tum.de:/FaceForensics/models/faceforensics++_models.zip).   
 
-Setup:
+Setup (requires Python 3):
+- Install `virtualenv`: 
+    ```shell
+    pip install virtualenv
+    ```
+- Create a Python virtual environment:
+    ```shell
+    virtualenv env
+    ```
+- Activate virtual environment:
+    1. Windows:
+    ```shell
+    cd venv\Scripts
+    activate
+    cd ..\..
+    ```
+    2. Lunix / Mac:
+    ```shell
+    source venv/bin/activate
+    ```
 - Install required modules via `requirement.txt` file
+- Download pre-trained models ([`wget`](http://gnuwin32.sourceforge.net/packages/wget.htm) allows easier download)
+    ```shell
+    wget -O model/xception-b5690688.pth http://data.lip6.fr/cadene/pretrainedmodels/xception-b5690688.pth
+    ```
 - Run detection from a single video file or folder with
-```shell
-python detect_from_video.py
--i <path to input video or folder of videos with extenstion '.mp4' or '.avi'>
--m <path to model file, default is imagenet model
--o <path to output folder, will contain output video(s)
-```  
+    ```shell
+    python detect_from_video.py
+    -i <path to input video or folder of videos with extenstion '.mp4' or '.avi'>
+    -m <path to model file, default is imagenet model
+    -o <path to output folder, will contain output video(s)
+    ```  
 from the classification folder. Enable cuda with ```--cuda```  or see parameters with ```python detect_from_video.py -h```.
 
 

classification/network/models.py

@@ -23,7 +23,7 @@ def return_pytorch04_xception(pretrained=True):
         model.fc = model.last_linear
         del model.last_linear
         state_dict = torch.load(
-            '/home/ondyari/.torch/models/xception-b5690688.pth')
+            'model/xception-b5690688.pth')
         for name, weights in state_dict.items():
             if 'pointwise' in name:
                 state_dict[name] = weights.unsqueeze(-1).unsqueeze(-1)

classification/requirements.txt

@@ -3,7 +3,7 @@ astor==0.7.1
 certifi==2018.11.29
 cffi==1.12.1
 cmake==3.12.0
-dlib==19.15.0
+dlib==19.7.0
 face-recognition==1.2.3
 face-recognition-models==0.3.0
 ffmpy==0.2.2
@@ -12,10 +12,10 @@ grpcio==1.14.1
 h5py==2.8.0
 Keras==2.2.0
 Markdown==2.6.11
-mkl-fft==1.0.10
-mkl-random==1.0.2
+mkl-fft
+mkl-random
 munch==2.3.2
-numpy==1.16.2
+numpy==1.18.1
 nvidia-ml-py3==7.352.0
 olefile==0.46
 opencv-python==3.4.1.15
@@ -27,6 +27,7 @@ pycparser==2.19
 scandir==1.7
 six==1.12.0
 termcolor==1.1.0
-torch==1.0.1.post2
-torchvision==0.2.1
+--find-links https://download.pytorch.org/whl/torch_stable.html
+torch==1.3.1
+torchvision==0.4.2
 tqdm==4.25.0