add READMEs for w2l

models/Wav2Lip/checkpoints/README.md

@@ -0,0 +1 @@
+Place all your checkpoints (.pth files) here. 

models/Wav2Lip/evaluation/README.md

@@ -0,0 +1,59 @@
+# Evaluation of Lip-sync using LSE-D and LSE-C metric.
+
+We use the pre-trained syncnet model available in this [repository](https://github.com/joonson/syncnet_python). 
+
+### Steps to set-up the evaluation repository for LSE-D and LSE-C metric:
+* Clone the SyncNet repository.
+``` 
+git clone https://github.com/joonson/syncnet_python.git 
+```
+* Follow the procedure given in the above linked [repository](https://github.com/joonson/syncnet_python) to download the pretrained models and set up the dependencies. 
+    * **Note: Please install a separate virtual environment for the evaluation scripts. The versions used by Wav2Lip and the publicly released code of SyncNet is different and can cause version mis-match issues. To avoid this, we suggest the users to install a separate virtual environment for the evaluation scripts**
+```
+cd syncnet_python
+pip install -r requirements.txt
+sh download_model.sh
+```
+* The above step should ensure that all the dependencies required by the repository is installed and the pre-trained models are downloaded.
+
+### Running the evaluation scripts:
+* Copy our evaluation scripts given in this folder to the cloned repository.
+```  
+    cd Wav2Lip/evaluation/scores_LSE/
+    cp *.py syncnet_python/
+    cp *.sh syncnet_python/ 
+```
+**Note: We will release the test filelists for LRW, LRS2 and LRS3 shortly once we receive permission from the dataset creators. We will also release the Real World Dataset we have collected shortly.**
+
+* Our evaluation technique does not require ground-truth of any sorts. Given lip-synced videos we can directly calculate the scores from only the generated videos. Please store the generated videos (from our test sets or your own generated videos) in the following folder structure.
+```
+video data root (Folder containing all videos)
+├── All .mp4 files
+```
+* Change the folder back to the cloned repository. 
+```
+cd syncnet_python
+```
+* To run evaluation on the LRW, LRS2 and LRS3 test files, please run the following command:
+```
+python calculate_scores_LRS.py --data_root /path/to/video/data/root --tmp_dir tmp_dir/
+```
+
+* To run evaluation on the ReSynced dataset or your own generated videos, please run the following command:
+```
+sh calculate_scores_real_videos.sh /path/to/video/data/root
+```
+* The generated scores will be present in the all_scores.txt generated in the ```syncnet_python/``` folder
+
+# Evaluation of image quality using FID metric.
+We will update the scripts for the FID metric shortly
+
+
+# Opening issues related to evaluation scripts
+* Please open the issues with the "Evaluation" label if you face any issues in the evaluation scripts. 
+
+# Acknowledgements
+Our evaluation pipeline in based on two existing repositories. LSE metrics are based on the [syncnet_python](https://github.com/joonson/syncnet_python) repository and the FID score is based on [pytorch-fid](https://github.com/mseitzer/pytorch-fid) repository. We thank the authors of both the repositories for releasing their wonderful code.
+
+
+

models/Wav2Lip/evaluation/test_filelists/README.md

@@ -0,0 +1 @@
+Place LRS2, LRW, and LRS3 (and any other) test set filelists here.

models/Wav2Lip/evaluation/test_filelists/ReSyncED/random_pairs.txt

@@ -0,0 +1,160 @@
+sachin.mp4 emma_cropped.mp4
+sachin.mp4 mourinho.mp4
+sachin.mp4 elon.mp4
+sachin.mp4 messi2.mp4
+sachin.mp4 cr1.mp4
+sachin.mp4 sachin.mp4
+sachin.mp4 sg.mp4
+sachin.mp4 fergi.mp4
+sachin.mp4 spanish_lec1.mp4
+sachin.mp4 bush_small.mp4
+sachin.mp4 macca_cut.mp4
+sachin.mp4 ca_cropped.mp4
+sachin.mp4 lecun.mp4
+sachin.mp4 spanish_lec0.mp4
+srk.mp4 emma_cropped.mp4
+srk.mp4 mourinho.mp4
+srk.mp4 elon.mp4
+srk.mp4 messi2.mp4
+srk.mp4 cr1.mp4
+srk.mp4 srk.mp4
+srk.mp4 sachin.mp4
+srk.mp4 sg.mp4
+srk.mp4 fergi.mp4
+srk.mp4 spanish_lec1.mp4
+srk.mp4 bush_small.mp4
+srk.mp4 macca_cut.mp4
+srk.mp4 ca_cropped.mp4
+srk.mp4 guardiola.mp4
+srk.mp4 lecun.mp4
+srk.mp4 spanish_lec0.mp4
+cr1.mp4 emma_cropped.mp4
+cr1.mp4 elon.mp4
+cr1.mp4 messi2.mp4
+cr1.mp4 cr1.mp4
+cr1.mp4 spanish_lec1.mp4
+cr1.mp4 bush_small.mp4
+cr1.mp4 macca_cut.mp4
+cr1.mp4 ca_cropped.mp4
+cr1.mp4 lecun.mp4
+cr1.mp4 spanish_lec0.mp4
+macca_cut.mp4 emma_cropped.mp4
+macca_cut.mp4 elon.mp4
+macca_cut.mp4 messi2.mp4
+macca_cut.mp4 spanish_lec1.mp4
+macca_cut.mp4 macca_cut.mp4
+macca_cut.mp4 ca_cropped.mp4
+macca_cut.mp4 spanish_lec0.mp4
+lecun.mp4 emma_cropped.mp4
+lecun.mp4 elon.mp4
+lecun.mp4 messi2.mp4
+lecun.mp4 spanish_lec1.mp4
+lecun.mp4 macca_cut.mp4
+lecun.mp4 ca_cropped.mp4
+lecun.mp4 lecun.mp4
+lecun.mp4 spanish_lec0.mp4
+messi2.mp4 emma_cropped.mp4
+messi2.mp4 elon.mp4
+messi2.mp4 messi2.mp4
+messi2.mp4 spanish_lec1.mp4
+messi2.mp4 macca_cut.mp4
+messi2.mp4 ca_cropped.mp4
+messi2.mp4 spanish_lec0.mp4
+ca_cropped.mp4 emma_cropped.mp4
+ca_cropped.mp4 elon.mp4
+ca_cropped.mp4 spanish_lec1.mp4
+ca_cropped.mp4 ca_cropped.mp4
+ca_cropped.mp4 spanish_lec0.mp4
+spanish_lec1.mp4 spanish_lec1.mp4
+spanish_lec1.mp4 spanish_lec0.mp4
+elon.mp4 elon.mp4
+elon.mp4 spanish_lec1.mp4
+elon.mp4 spanish_lec0.mp4
+guardiola.mp4 emma_cropped.mp4
+guardiola.mp4 mourinho.mp4
+guardiola.mp4 elon.mp4
+guardiola.mp4 messi2.mp4
+guardiola.mp4 cr1.mp4
+guardiola.mp4 sachin.mp4
+guardiola.mp4 sg.mp4
+guardiola.mp4 fergi.mp4
+guardiola.mp4 spanish_lec1.mp4
+guardiola.mp4 bush_small.mp4
+guardiola.mp4 macca_cut.mp4
+guardiola.mp4 ca_cropped.mp4
+guardiola.mp4 guardiola.mp4
+guardiola.mp4 lecun.mp4
+guardiola.mp4 spanish_lec0.mp4
+fergi.mp4 emma_cropped.mp4
+fergi.mp4 mourinho.mp4
+fergi.mp4 elon.mp4
+fergi.mp4 messi2.mp4
+fergi.mp4 cr1.mp4
+fergi.mp4 sachin.mp4
+fergi.mp4 sg.mp4
+fergi.mp4 fergi.mp4
+fergi.mp4 spanish_lec1.mp4
+fergi.mp4 bush_small.mp4
+fergi.mp4 macca_cut.mp4
+fergi.mp4 ca_cropped.mp4
+fergi.mp4 lecun.mp4
+fergi.mp4 spanish_lec0.mp4
+spanish.mp4 emma_cropped.mp4
+spanish.mp4 spanish.mp4
+spanish.mp4 mourinho.mp4
+spanish.mp4 elon.mp4
+spanish.mp4 messi2.mp4
+spanish.mp4 cr1.mp4
+spanish.mp4 srk.mp4
+spanish.mp4 sachin.mp4
+spanish.mp4 sg.mp4
+spanish.mp4 fergi.mp4
+spanish.mp4 spanish_lec1.mp4
+spanish.mp4 bush_small.mp4
+spanish.mp4 macca_cut.mp4
+spanish.mp4 ca_cropped.mp4
+spanish.mp4 guardiola.mp4
+spanish.mp4 lecun.mp4
+spanish.mp4 spanish_lec0.mp4
+bush_small.mp4 emma_cropped.mp4
+bush_small.mp4 elon.mp4
+bush_small.mp4 messi2.mp4
+bush_small.mp4 spanish_lec1.mp4
+bush_small.mp4 bush_small.mp4
+bush_small.mp4 macca_cut.mp4
+bush_small.mp4 ca_cropped.mp4
+bush_small.mp4 lecun.mp4
+bush_small.mp4 spanish_lec0.mp4
+emma_cropped.mp4 emma_cropped.mp4
+emma_cropped.mp4 elon.mp4
+emma_cropped.mp4 spanish_lec1.mp4
+emma_cropped.mp4 spanish_lec0.mp4
+sg.mp4 emma_cropped.mp4
+sg.mp4 mourinho.mp4
+sg.mp4 elon.mp4
+sg.mp4 messi2.mp4
+sg.mp4 cr1.mp4
+sg.mp4 sachin.mp4
+sg.mp4 sg.mp4
+sg.mp4 fergi.mp4
+sg.mp4 spanish_lec1.mp4
+sg.mp4 bush_small.mp4
+sg.mp4 macca_cut.mp4
+sg.mp4 ca_cropped.mp4
+sg.mp4 lecun.mp4
+sg.mp4 spanish_lec0.mp4
+spanish_lec0.mp4 spanish_lec0.mp4
+mourinho.mp4 emma_cropped.mp4
+mourinho.mp4 mourinho.mp4
+mourinho.mp4 elon.mp4
+mourinho.mp4 messi2.mp4
+mourinho.mp4 cr1.mp4
+mourinho.mp4 sachin.mp4
+mourinho.mp4 sg.mp4
+mourinho.mp4 fergi.mp4
+mourinho.mp4 spanish_lec1.mp4
+mourinho.mp4 bush_small.mp4
+mourinho.mp4 macca_cut.mp4
+mourinho.mp4 ca_cropped.mp4
+mourinho.mp4 lecun.mp4
+mourinho.mp4 spanish_lec0.mp4

models/Wav2Lip/evaluation/test_filelists/ReSyncED/tts_pairs.txt

@@ -0,0 +1,18 @@
+adam_1.mp4 andreng_optimization.wav
+agad_2.mp4 agad_2.wav
+agad_1.mp4 agad_1.wav
+agad_3.mp4 agad_3.wav
+rms_prop_1.mp4 rms_prop_tts.wav
+tf_1.mp4 tf_1.wav
+tf_2.mp4 tf_2.wav
+andrew_ng_ai_business.mp4 andrewng_business_tts.wav
+covid_autopsy_1.mp4 autopsy_tts.wav
+news_1.mp4 news_tts.wav
+andrew_ng_fund_1.mp4 andrewng_ai_fund.wav
+covid_treatments_1.mp4 covid_tts.wav
+pytorch_v_tf.mp4 pytorch_vs_tf_eng.wav
+pytorch_1.mp4 pytorch.wav
+pkb_1.mp4 pkb_1.wav
+ss_1.mp4 ss_1.wav
+carlsen_1.mp4 carlsen_eng.wav
+french.mp4 french.wav

models/Wav2Lip/face_detection/README.md

@@ -0,0 +1 @@
+The code for Face Detection in this folder has been taken from the wonderful [face_alignment](https://github.com/1adrianb/face-alignment) repository. This has been modified to take batches of faces at a time. 

models/Wav2Lip/filelists/README.md

@@ -0,0 +1 @@
+Place LRS2 (and any other) filelists here for training.

models/Wav2Lip/results/README.md

@@ -0,0 +1 @@
+Generated results will be placed in this folder by default.

models/Wav2Lip/temp/README.md

@@ -0,0 +1 @@
+Temporary files at the time of inference/testing will be saved here. You can ignore them.