Merge branch 'JVET-AB0053_training' into 'VTM-11.0_nnvc'

JVET-AB0053: EE1-1.2: (second merge) Incorporates training scripts for the luma intra model for NN filter set 1.

See merge request jvet-ahg-nnvc/VVCSoftware_VTM!23

training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/README_training_luma_intra.txt

+
+Step A) Checkout the training code:
+===================================
+
+git clone https://vcgit.hhi.fraunhofer.de/jvet-ahg-nnvc/VVCSoftware_VTM.git
+cd VVCSoftware_VTM
+git checkout NNVC-3.0
+This is now your $BASE path
+
+cd $BASE/training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/
+This is now your $TRAIN path
+
+cd $BASE/training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/data_extraction_scripts/
+This is now your $EXTRACT path
+
+
+
+Step B) Download ffmpeg:
+========================
+
+wget https://johnvansickle.com/ffmpeg/releases/ffmpeg-release-amd64-static.tar.xz
+tar xvf ffmpeg-release-amd64-static.tar.xz
+cd ffmpeg-5.0.1-amd64-static
+pwd
+
+--> This is now your path 'ffmpeg_path' in $EXTRACT/png2yuv_subsampled.py and $EXTRACT/png2yuv_fullres.py
+--> It is also your path 'executablefile' in $EXTRACT/mp42yuv_BVI-DVC.py
+
+
+
+Step C) Prepare subsampled yuv-files for DIV2K
+==============================================
+
+Goto https://data.vision.ee.ethz.ch/cvl/DIV2K/
+Download half resolution images from http://data.vision.ee.ethz.ch/cvl/DIV2K/DIV2K_train_LR_bicubic_X2.zip
+unzip DIV2K_train_LR_bicubic_X2.zip
+cd DIV2K_train_LR_bicubic/X2
+pwd
+--> This is now your path 'srcdir' in $EXTRACT/png2yuv_subsampled.py
+(Mine: /proj/video_no_backup/videosim/ejacstr/extract/DIV2K_test_yuv/DIV2K_train_LR_bicubic/X2)
+cd ..
+mkdir X2_crop
+cd X2_crop
+pwd
+--> This is now your path 'cropimgdir' in $EXTRACT/png2yuv_subsampled.py
+(Mine: /proj/video_no_backup/videosim/ejacstr/extract/DIV2K_test_yuv/X2_crop)
+cd ..
+mkdir X2_yuv
+cd X2_yuv
+pwd
+--> This is now your path 'destdir' in $EXTRACT/png2yuv_subsampled.py, 'inputdirDIV2K_X2' in $EXTRACT/E_data_generation_luma_intra.py and 'path_to_original_DIV2K_X2_yuvs' in $TRAIN/main.py
+
+After changing all the paths in png2yuv_subsampled.py according to the above, run png2yuv_subsampled.py using
+cd $EXTRACT
+python3 png2yuv_subsampled.py
+
+The subsampled versions of DIV2K images will afterwards reside in 'destdir' of the form
+
+0084x2_1016x680_25fps_8bit_420.yuv
+signifying image 84, downsampled a factor of two (x2), resolution 1016x680, 25 fps, 8bit 420.
+
+
+
+Step D) Prepare full resolution yuv-files for DIV2K:
+===============================================
+
+Goto https://data.vision.ee.ethz.ch/cvl/DIV2K/
+Download full resolution images from http://data.vision.ee.ethz.ch/cvl/DIV2K/DIV2K_train_HR.zip
+unzip DIV2K_train_HR.zip
+cd DIV2K_train_HR
+pwd
+--> This is now your path 'srcdir' in $EXTRACT/png2yuv_fullres.py
+cd ..
+mkdir DIV2K_train_HR_crop
+cd DIV2K_train_HR_crop
+pwd
+--> This is now your path 'cropimgdir' in $EXTRACT/png2yuv_fullres.py
+cd ..
+mkdir DIV2K_train_HR_yuv
+cd DIV2K_train_HR_yuv
+pwd
+--> This is now your path 'destdir' in $EXTRACT/png2yuv_fullres.py, 'inputdirDIV2K' in $EXTRACT/E_data_generation_luma_intra.py and 'path_to_original_DIV2K_yuvs' in $TRAIN/main.py.
+
+After changing all the paths in png2yuv_fullres.py according to the above, run png2yuv_fullres.py using
+cd $EXTRACT
+python3 png2yuv_fullres.py
+
+The full resolution yuv versions of the DIV2K images will afterwards reside in 'destdir' of the form
+
+0461_1440x2040_25fps_8bit_420.yuv
+signifying image 461, no downsampling (lack of x2), resolution 1440x2040, 25 fps, 8bit 420.
+
+E) Prepare yuv-files for BVI-DVC:
+=================================
+
+Download the BVI-DVC dataset as per the instructions in the NNVC-CTC.
+--> The path where you have downloaded all the mp4 files is now your 'srcdir' in mp42yuv_BVI-DVC.py
+
+Find a suitable path to put your .yuv videos
+--> This is now your 'destdir' in $EXTRACT/mp42yuv_BVI-DVC.py, 'inputdirBVIDVC' in $EXTRACT/E_data_generation_luma_intra.py and 'path_to_original_BVI_DVC_yuvs' in $TRAIN/main.py
+
+--> Change 'mycurrentdir' to the directory where the script file mp42yuv_BVI-DVC.py is residing.
+
+--> Change the path 'executablefile' in $EXTRACT/mp42yuv_BVI-DVC.py to the ffmpeg path you created above.
+
+Run mp42yuv_BVI-DVC.py four times for the different sizes A (), B (1920x1088), C (960x544) and D (480x272):
+
+cd $EXTRACT
+python3 mp42yuv_BVI-DVC.py A
+python3 mp42yuv_BVI-DVC.py B
+python3 mp42yuv_BVI-DVC.py C
+python3 mp42yuv_BVI-DVC.py D
+
+Note that each of these scripts starts 192 several simultaneous decodings. If your system cannot cope with that many simultaneous decodings, use something like time.sleep(20) to wait 20 seconds after starting a decoding instead of doing it instantly.
+
+to save your changes in the above scripts, you can create a new branch>
+
+cd $BASE
+git checkout -b scripts_with_paths
+git commit -am "Scripts updated with my paths."
+
+
+F) Run NNVC-VTM for DIV2K, DIV2K-x2 and BVI-DVC and generate training data:
+===========================================================================
+
+F1) Compile NNVC-encoder and extractor decoder:
+-----------------------------------------------
+
+We use NCS-1.0 as extraction software. It is possible to use a later version, but this has not been tested to give equivalent results.
+NCS-1.0 is equivalent to commit b4df98075df77f779a86b4b76a6bcfabf9d6fecd. Therefore,
+
+cd $BASE
+git checkout b4df98075df77f779a86b4b76a6bcfabf9d6fecd
+git checkout -b extraction_software
+
+In source/Lib/CommonLib/TypeDef.h, set the macro NNVC_USE_SLICETYPE to 0:
+#define NNVC_USE_SLICETYPE                                0 // slice type 
+
+Commit your changes to the new branch
+
+git commit -am "Extraction software now ready."
+
+Now compile, first by updating SADL
+
+git submodule init
+git submodule update
+
+Then build:
+
+mkdir build
+cd build
+cmake .. -DCMAKE_BUILD_TYPE=Release
+make -j 8
+cd ../bin
+cp EncoderAppStatic EncoderAppStatic_extraction_software
+cp DecoderAppStatic DecoderAppStatic_extraction_software
+pwd
+
+Copy the resulting path to the variable 'excutabledir' in E_data_generation_luma_intra.py. Before updating the file with this path, checkout your updated scripts with
+
+git checkout scripts_with_paths
+
+and then edit $EXTRACT/E_data_generation_luma_intra.py.
+
+
+F2) Compress yuv-files and extract training data
+------------------------------------------------
+
+In $EXTRACT/E_data_generation_luma_intra.py, change the following:
+
+Change 'bitstreamsdir' to a directory on your harddrive where you want to store the resulting bitstreams.
+Change 'dumpdirbase' to a directory where you want the decoded yuv data. Also change 'default_base_train_dir' in $TRAIN/main.py to this directory.
+Change 'mycurrentdir' to the full path for the directory where the script file E_data_generation_luma_intra.py is residing.
+
+The paths 'inputdirDIV2K', 'inputdirDIV2K_X2' and 'inputdirBVIDVC' were updated in previous steps and need not be futher changed. 
+
+You may want to change the line
+
+cmdall = ["bsub -o /dev/null '", enccmd,";",deccmd, "'"]
+
+to something that starts jobs our your cluster. We are using LSF and start a job with " bsub -o /dev/null 'python3 jobtostart.py' ".
+
+For DIV2K:
+
+cd $EXTRACT
+python3 E_data_generation_luma_intra.py DIV2K 0
+
+For subsampled DIV2K:
+
+python3 E_data_generation_luma_intra.py DIV2K_X2 0
+
+For BVIDVC:
+
+python3 E_data_generation_luma_intra.py BVIDVC A
+python3 E_data_generation_luma_intra.py BVIDVC B
+python3 E_data_generation_luma_intra.py BVIDVC C
+python3 E_data_generation_luma_intra.py BVIDVC D
+
+The resulting yuv-files to be used for training will now reside in 'dumpdirbase'
+
+
+
+G) Train simplified model:
+==========================
+Start training using
+cd $TRAIN
+python3 main.py --learning_rate 1e-4 --loss_function 'L1' --num_workers 32 --epochs 371  --tag simple_model
+
+

training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/data_extraction_scripts/E_data_generation_luma_intra.py

+import csv
+import sys
+import os
+from sys import platform
+import subprocess
+from subprocess import Popen, PIPE
+import shutil
+import glob
+import time
+
+
+cfgtype = 'ai' 
+
+
+# A list of QPs you want to encode. The submitted jobs will loop all the QPs. Remove the final two "break"s when ready to submit coding jobs.
+#QP = [45,40,35,30,25,20]
+QP = [20, 25, 30, 35, 40]
+
+
+#dataset is either 'DIV2K', 'DIV2K_X2', or 'BVIDVC'
+#dataset = 'BVIDVC'
+#dataset = 'DIV2K_X2'
+dataset = sys.argv[1]
+
+# String to control which names from the dataset you want to encode. 
+# E.g., DIV2K images start with 0xxx_xxxxxxx_xxx.yuv, so str1='0'. 
+# BVIDVC class D sequences start with Dxxxx_xxxxx.yuv, so str1='D'.
+# A paticular sequence with name CColourfulDecorationWatPhoVidevo_960x544_50fps_10bit_420.yuv, so str1 = 'CColourfulDecorationWatPhoVidevo_960x544_50fps_10bit_420'
+#str1 = 'DAdvertisingMassagesBangkokVidevo_480x272_25fps_10bit_420'
+#str1 = 'AWovenVidevo_3840x2176_25fps_10bit_420'
+
+str1 = sys.argv[2]    
+
+###############
+# Change the following directories to your directories
+
+# directory of encoder and decoder
+excutabledir = '/proj/video_no_backup/videosim/ejacstr/jvet_training_extraction/jvet_training/bin/'
+encoderdir = excutabledir+'EncoderAppStatic_extraction_software'
+decoderdir = excutabledir+'DecoderAppStatic_extraction_software'
+
+# directory where you can find original DIV2K yuv files of full resolution
+inputdirDIV2K = '/proj/video_no_backup/videosim/ejacstr/jvet_training_extraction/links_to_AB0053_data/DIV2K_train_HR_yuv/'
+# directory where you can find original DIV2K yuv files of half resolution
+inputdirDIV2K_X2 = '/proj/video_no_backup/videosim/ejacstr/jvet_training_extraction/links_to_AB0053_data/DIV2K_train_X2_yuv/'
+# directory where you can find original BVIDVC yuv files
+inputdirBVIDVC = '/proj/video_no_backup/videosim/ejacstr/jvet_training_extraction/links_to_AB0053_data/BVI_DVC_train_yuv/'
+
+# Output directory of compressed bitstreams. Change the first part of the path to a location of your choice.
+bitstreamsdir = '/proj/video_no_backup/videosim/ejacstr/jvet_training_extraction/links_to_AB0053_data/bitstreams/'
+yuvoutputdir = bitstreamsdir + cfgtype + '_train/'
+enclog_outputdir = bitstreamsdir + cfgtype + '_train_enclogs/'
+# Dump directory where decompressed yuv files used for training data will reside. 
+dumpdirbase = '/proj/video_no_backup/videosim/ejacstr/jvet_training_extraction/links_to_AB0053_data/train_dump/'
+os.makedirs(yuvoutputdir, exist_ok=True)
+os.makedirs(enclog_outputdir, exist_ok=True)
+os.makedirs(dumpdirbase, exist_ok=True)
+
+# Current directory (where this script file resides).
+mycurrentdir = '/proj/video_no_backup/videosim/ejacstr/jvet_training_extraction/jvet_training/training/training_scripts/Nn_Filtering_Set_1/Scripts/data_extraction_scripts/'
+
+
+################
+
+#AI configuration file
+cfg1 = mycurrentdir + '/encoder_intra_vtm_all.cfg'
+
+# Output directory of final training data. If necessary, change to a proper directory.
+outputroot = "/path/to/train/"+cfgtype+"/"
+
+###############
+    
+if dataset == 'DIV2K':
+    cfgdir = mycurrentdir + '/cfgDIV2K_IntraLuma/'
+    datasetname = "DIV2K_1frm" 
+    inputdir = inputdirDIV2K
+    encnumfrm = '1'  #number of frames to encode
+    numfrm = '1'  # total number of frames
+    tempSubSampling = '1'    
+    bitdepth = 10
+    bitdepth_orig = 8 # BVIDVC 10bit, DIV2K 8bit        
+    logdir = ''# not needed for AI
+elif dataset == 'DIV2K_X2':
+    datasetname = "DIV2K_X2_1frm"
+    cfgdir = mycurrentdir + '/cfgDIV2K_X2_IntraLuma/'
+    inputdir = inputdirDIV2K_X2
+    encnumfrm = '1'
+    numfrm = '1'
+    tempSubSampling = '1'
+    bitdepth = 10
+    bitdepth_orig = 8 # BVIDVC 10bit, DIV2K 8bit        
+    logdir = '' # not needed for AI
+elif dataset == 'BVIDVC':    
+    datasetname = "BVIDVC_64frm"
+    cfgdir = mycurrentdir + '/cfgBVIDVC_IntraLuma/'
+    inputdir =  inputdirBVIDVC
+    encnumfrm = '64'
+    numfrm = '64'
+    if str1[0] == 'A':
+        tempSubSampling = '8'
+    else:
+        tempSubSampling = '1'        
+    bitdepth = 10
+    bitdepth_orig = 10 # BVIDVC 10bit, DIV2K 8bit
+    logdir = ''# not needed for AI
+else:
+    print('dataset must be DIV2K_HR_and_XR. Exiting.')
+    exit(1)
+
+cnt = 0
+
+for input_file in os.listdir(inputdir):
+    print('input file = ', input_file)
+    #input_file = '0800_downscaling.yuv'
+    if input_file[-3:] == 'yuv' and input_file[:len(str1)] == str1:
+        seqname = input_file[:-4]
+        cfg2 = cfgdir + seqname + '.cfg'
+        file_dir = inputdir + input_file
+        dumpdir = dumpdirbase + seqname
+        os.makedirs(dumpdir, exist_ok=True)
+        for q in range(0,len(QP)):
+            qp = str(QP[q])
+            seqname_with_qp = yuvoutputdir + seqname + '_qp' + qp;
+            
+            bin_dir = seqname_with_qp + '_str.bin'
+            tmpout_dir = seqname_with_qp + '_tmprec.yuv'
+            log_dir = seqname_with_qp + '_dec.log'
+            enclog_dir = enclog_outputdir + seqname + '_qp' + qp + '_enc.log'
+            
+            enccmd = [encoderdir, "-c", cfg1, "-c", cfg2, "-i", file_dir, "-o", tmpout_dir,"-b", bin_dir, "-q", qp, "-f",encnumfrm, "-ts",tempSubSampling, "--NnlfOption=0 --Verbosity=6", ">&",enclog_dir]
+            enccmd = " ".join(enccmd) 
+            
+            deccmd = [decoderdir, "-b", bin_dir, "--DumpBasename="+dumpdir+'/'+seqname+"_qp"+qp,">&",log_dir]
+            deccmd = " ".join(deccmd)
+            
+            cmdall = ["bsub -o /dev/null '", enccmd,";",deccmd, "'"]
+
+            cmdall = " ".join(cmdall)
+            print(cmdall)
+            os.system(cmdall)
+            cnt += 1 
+            time.sleep(0.3)
+            
+print('Total',cnt,'jobs.')
+
+# Unresolved: What main cfg file is used?

training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/data_extraction_scripts/cfgBVIDVC_IntraLuma/AAdvertisingMassagesBangkokVidevo_3840x2176_25fps_10bit_420.cfg

+#======== File I/O ===============
+InputFile                     : AAdvertisingMassagesBangkokVidevo_3840x2176_25fps_10bit_420.yuv
+
+InputBitDepth                 : 10           # Input bitdepth
+InputChromaFormat             : 420         # Ratio of luminance to chrominance samples
+FrameRate                     : 25          # Frame Rate per second
+FrameSkip                     : 0           # Number of frames to be skipped in input
+SourceWidth                   : 3840         # Input  frame width
+SourceHeight                  : 2176         # Input  frame height
+FramesToBeEncoded             : 64         # Number of frames to be coded
+
+Level                         : 5.1

training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/data_extraction_scripts/cfgBVIDVC_IntraLuma/AAmericanFootballS2Harmonics_3840x2176_60fps_10bit_420.cfg

+#======== File I/O ===============
+InputFile                     : AAmericanFootballS2Harmonics_3840x2176_60fps_10bit_420.yuv
+
+InputBitDepth                 : 10           # Input bitdepth
+InputChromaFormat             : 420         # Ratio of luminance to chrominance samples
+FrameRate                     : 60          # Frame Rate per second
+FrameSkip                     : 0           # Number of frames to be skipped in input
+SourceWidth                   : 3840         # Input  frame width
+SourceHeight                  : 2176         # Input  frame height
+FramesToBeEncoded             : 64         # Number of frames to be coded
+
+Level                         : 5.1

training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/data_extraction_scripts/cfgBVIDVC_IntraLuma/AAmericanFootballS3Harmonics_3840x2176_60fps_10bit_420.cfg

+#======== File I/O ===============
+InputFile                     : AAmericanFootballS3Harmonics_3840x2176_60fps_10bit_420.yuv
+
+InputBitDepth                 : 10           # Input bitdepth
+InputChromaFormat             : 420         # Ratio of luminance to chrominance samples
+FrameRate                     : 60          # Frame Rate per second
+FrameSkip                     : 0           # Number of frames to be skipped in input
+SourceWidth                   : 3840         # Input  frame width
+SourceHeight                  : 2176         # Input  frame height
+FramesToBeEncoded             : 64         # Number of frames to be coded
+
+Level                         : 5.1

training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/data_extraction_scripts/cfgBVIDVC_IntraLuma/AAmericanFootballS4Harmonics_3840x2176_60fps_10bit_420.cfg

+#======== File I/O ===============
+InputFile                     : AAmericanFootballS4Harmonics_3840x2176_60fps_10bit_420.yuv
+
+InputBitDepth                 : 10           # Input bitdepth
+InputChromaFormat             : 420         # Ratio of luminance to chrominance samples
+FrameRate                     : 60          # Frame Rate per second
+FrameSkip                     : 0           # Number of frames to be skipped in input
+SourceWidth                   : 3840         # Input  frame width
+SourceHeight                  : 2176         # Input  frame height
+FramesToBeEncoded             : 64         # Number of frames to be coded
+
+Level                         : 5.1

training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/data_extraction_scripts/cfgBVIDVC_IntraLuma/AAnimalsS11Harmonics_3840x2176_60fps_10bit_420.cfg

+#======== File I/O ===============
+InputFile                     : AAnimalsS11Harmonics_3840x2176_60fps_10bit_420.yuv
+
+InputBitDepth                 : 10           # Input bitdepth
+InputChromaFormat             : 420         # Ratio of luminance to chrominance samples
+FrameRate                     : 60          # Frame Rate per second
+FrameSkip                     : 0           # Number of frames to be skipped in input
+SourceWidth                   : 3840         # Input  frame width
+SourceHeight                  : 2176         # Input  frame height
+FramesToBeEncoded             : 64         # Number of frames to be coded
+
+Level                         : 5.1

training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/data_extraction_scripts/cfgBVIDVC_IntraLuma/AAnimalsS1Harmonics_3840x2176_60fps_10bit_420.cfg

+#======== File I/O ===============
+InputFile                     : AAnimalsS1Harmonics_3840x2176_60fps_10bit_420.yuv
+
+InputBitDepth                 : 10           # Input bitdepth
+InputChromaFormat             : 420         # Ratio of luminance to chrominance samples
+FrameRate                     : 60          # Frame Rate per second
+FrameSkip                     : 0           # Number of frames to be skipped in input
+SourceWidth                   : 3840         # Input  frame width
+SourceHeight                  : 2176         # Input  frame height
+FramesToBeEncoded             : 64         # Number of frames to be coded
+
+Level                         : 5.1

training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/data_extraction_scripts/cfgBVIDVC_IntraLuma/ABangkokMarketVidevo_3840x2176_25fps_10bit_420.cfg

+#======== File I/O ===============
+InputFile                     : ABangkokMarketVidevo_3840x2176_25fps_10bit_420.yuv
+
+InputBitDepth                 : 10           # Input bitdepth
+InputChromaFormat             : 420         # Ratio of luminance to chrominance samples
+FrameRate                     : 25          # Frame Rate per second
+FrameSkip                     : 0           # Number of frames to be skipped in input
+SourceWidth                   : 3840         # Input  frame width
+SourceHeight                  : 2176         # Input  frame height
+FramesToBeEncoded             : 64         # Number of frames to be coded
+
+Level                         : 5.1

training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/data_extraction_scripts/cfgBVIDVC_IntraLuma/ABasketballGoalScoredS1Videvo_3840x2176_25fps_10bit_420.cfg

+#======== File I/O ===============
+InputFile                     : ABasketballGoalScoredS1Videvo_3840x2176_25fps_10bit_420.yuv
+
+InputBitDepth                 : 10           # Input bitdepth
+InputChromaFormat             : 420         # Ratio of luminance to chrominance samples
+FrameRate                     : 25          # Frame Rate per second
+FrameSkip                     : 0           # Number of frames to be skipped in input
+SourceWidth                   : 3840         # Input  frame width
+SourceHeight                  : 2176         # Input  frame height
+FramesToBeEncoded             : 64         # Number of frames to be coded
+
+Level                         : 5.1

training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/data_extraction_scripts/cfgBVIDVC_IntraLuma/ABasketballGoalScoredS2Videvo_3840x2176_25fps_10bit_420.cfg

+#======== File I/O ===============
+InputFile                     : ABasketballGoalScoredS2Videvo_3840x2176_25fps_10bit_420.yuv
+
+InputBitDepth                 : 10           # Input bitdepth
+InputChromaFormat             : 420         # Ratio of luminance to chrominance samples
+FrameRate                     : 25          # Frame Rate per second
+FrameSkip                     : 0           # Number of frames to be skipped in input
+SourceWidth                   : 3840         # Input  frame width
+SourceHeight                  : 2176         # Input  frame height
+FramesToBeEncoded             : 64         # Number of frames to be coded
+
+Level                         : 5.1

training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/data_extraction_scripts/cfgBVIDVC_IntraLuma/ABasketballS1YonseiUniversity_3840x2176_30fps_10bit_420.cfg

+#======== File I/O ===============
+InputFile                     : ABasketballS1YonseiUniversity_3840x2176_30fps_10bit_420.yuv
+
+InputBitDepth                 : 10           # Input bitdepth
+InputChromaFormat             : 420         # Ratio of luminance to chrominance samples
+FrameRate                     : 30          # Frame Rate per second
+FrameSkip                     : 0           # Number of frames to be skipped in input
+SourceWidth                   : 3840         # Input  frame width
+SourceHeight                  : 2176         # Input  frame height
+FramesToBeEncoded             : 64         # Number of frames to be coded
+
+Level                         : 5.1

training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/data_extraction_scripts/cfgBVIDVC_IntraLuma/ABasketballS2YonseiUniversity_3840x2176_30fps_10bit_420.cfg

+#======== File I/O ===============
+InputFile                     : ABasketballS2YonseiUniversity_3840x2176_30fps_10bit_420.yuv
+
+InputBitDepth                 : 10           # Input bitdepth
+InputChromaFormat             : 420         # Ratio of luminance to chrominance samples
+FrameRate                     : 30          # Frame Rate per second
+FrameSkip                     : 0           # Number of frames to be skipped in input
+SourceWidth                   : 3840         # Input  frame width
+SourceHeight                  : 2176         # Input  frame height
+FramesToBeEncoded             : 64         # Number of frames to be coded
+
+Level                         : 5.1

training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/data_extraction_scripts/cfgBVIDVC_IntraLuma/ABasketballS3YonseiUniversity_3840x2176_30fps_10bit_420.cfg

+#======== File I/O ===============
+InputFile                     : ABasketballS3YonseiUniversity_3840x2176_30fps_10bit_420.yuv
+
+InputBitDepth                 : 10           # Input bitdepth
+InputChromaFormat             : 420         # Ratio of luminance to chrominance samples
+FrameRate                     : 30          # Frame Rate per second
+FrameSkip                     : 0           # Number of frames to be skipped in input
+SourceWidth                   : 3840         # Input  frame width
+SourceHeight                  : 2176         # Input  frame height
+FramesToBeEncoded             : 64         # Number of frames to be coded
+
+Level                         : 5.1

training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/data_extraction_scripts/cfgBVIDVC_IntraLuma/ABoatsChaoPhrayaRiverVidevo_3840x2176_23fps_10bit_420.cfg

+#======== File I/O ===============
+InputFile                     : ABoatsChaoPhrayaRiverVidevo_3840x2176_23fps_10bit_420.yuv
+
+InputBitDepth                 : 10           # Input bitdepth
+InputChromaFormat             : 420         # Ratio of luminance to chrominance samples
+FrameRate                     : 23          # Frame Rate per second
+FrameSkip                     : 0           # Number of frames to be skipped in input
+SourceWidth                   : 3840         # Input  frame width
+SourceHeight                  : 2176         # Input  frame height
+FramesToBeEncoded             : 64         # Number of frames to be coded
+
+Level                         : 5.1

training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/data_extraction_scripts/cfgBVIDVC_IntraLuma/ABobbleheadBVIHFR_3840x2176_120fps_10bit_420.cfg

+#======== File I/O ===============
+InputFile                     : ABobbleheadBVIHFR_3840x2176_120fps_10bit_420.yuv
+
+InputBitDepth                 : 10           # Input bitdepth
+InputChromaFormat             : 420         # Ratio of luminance to chrominance samples
+FrameRate                     : 120          # Frame Rate per second
+FrameSkip                     : 0           # Number of frames to be skipped in input
+SourceWidth                   : 3840         # Input  frame width
+SourceHeight                  : 2176         # Input  frame height
+FramesToBeEncoded             : 64         # Number of frames to be coded
+
+Level                         : 5.1

training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/data_extraction_scripts/cfgBVIDVC_IntraLuma/ABookcaseBVITexture_3840x2176_120fps_10bit_420.cfg

+#======== File I/O ===============
+InputFile                     : ABookcaseBVITexture_3840x2176_120fps_10bit_420.yuv
+
+InputBitDepth                 : 10           # Input bitdepth
+InputChromaFormat             : 420         # Ratio of luminance to chrominance samples
+FrameRate                     : 120          # Frame Rate per second
+FrameSkip                     : 0           # Number of frames to be skipped in input
+SourceWidth                   : 3840         # Input  frame width
+SourceHeight                  : 2176         # Input  frame height
+FramesToBeEncoded             : 64         # Number of frames to be coded
+
+Level                         : 5.1

training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/data_extraction_scripts/cfgBVIDVC_IntraLuma/ABoxingPracticeHarmonics_3840x2176_60fps_10bit_420.cfg

+#======== File I/O ===============
+InputFile                     : ABoxingPracticeHarmonics_3840x2176_60fps_10bit_420.yuv
+
+InputBitDepth                 : 10           # Input bitdepth
+InputChromaFormat             : 420         # Ratio of luminance to chrominance samples
+FrameRate                     : 60          # Frame Rate per second
+FrameSkip                     : 0           # Number of frames to be skipped in input
+SourceWidth                   : 3840         # Input  frame width
+SourceHeight                  : 2176         # Input  frame height
+FramesToBeEncoded             : 64         # Number of frames to be coded
+
+Level                         : 5.1

training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/data_extraction_scripts/cfgBVIDVC_IntraLuma/ABricksBushesStaticBVITexture_3840x2176_120fps_10bit_420.cfg

+#======== File I/O ===============
+InputFile                     : ABricksBushesStaticBVITexture_3840x2176_120fps_10bit_420.yuv
+
+InputBitDepth                 : 10           # Input bitdepth
+InputChromaFormat             : 420         # Ratio of luminance to chrominance samples
+FrameRate                     : 120          # Frame Rate per second
+FrameSkip                     : 0           # Number of frames to be skipped in input
+SourceWidth                   : 3840         # Input  frame width
+SourceHeight                  : 2176         # Input  frame height
+FramesToBeEncoded             : 64         # Number of frames to be coded
+
+Level                         : 5.1

training/training_scripts/Nn_Filtering_Set_1/Scripts/LumaIntra/data_extraction_scripts/cfgBVIDVC_IntraLuma/ABricksLeavesBVITexture_3840x2176_120fps_10bit_420.cfg

+#======== File I/O ===============
+InputFile                     : ABricksLeavesBVITexture_3840x2176_120fps_10bit_420.yuv
+
+InputBitDepth                 : 10           # Input bitdepth
+InputChromaFormat             : 420         # Ratio of luminance to chrominance samples
+FrameRate                     : 120          # Frame Rate per second
+FrameSkip                     : 0           # Number of frames to be skipped in input
+SourceWidth                   : 3840         # Input  frame width
+SourceHeight                  : 2176         # Input  frame height
+FramesToBeEncoded             : 64         # Number of frames to be coded
+
+Level                         : 5.1