SQL Database Content¶

For a more detailed analysis of constraints and relationships, see the contents of the function mendevi.database.create.create_database().

Table t_act_activity¶

act_id	INTEGER	key
act_start	TIMESTAMP	date when measures start
act_duration	FLOAT	duration of measures
act_rapl_dt	LONGBLOB	RAPL measures intervals
act_rapl_power	LONGBLOB	RAPL power values
act_wattmeter_dt	LONGBLOB	wattmeter measures intervals
act_wattmeter_power	LONGBLOB	wattmeter power values
act_gpu_dt	LONGBLOB	GPU measures intervals
act_gpu_power	LONGBLOB	GPU power values
act_ps_dt	LONGBLOB	time intervals for ps
act_ps_core	LONGBLOB	core usage values
act_ps_temp	JSON	temperature of all devices
act_ps_ram	LONGBLOB	RAM usage values

Captures machine activity over a period of time.

act_id¶

Unique primary key, with no particular meaning.

act_start¶

The date when the measure starts, the number of second from the 1st january 1970 at 00:00:00 UTC.

act_duration¶

The duration of the measurement in seconds.

act_rapl_dt¶

The list of each rapl measures duration in seconds. To deserialize into an exploitable numpy array, please use the function mendevi.database.serialize.binary_to_list().

act_rapl_power¶

List of average power measurement in w with the wattmeter directly integrated into the processor power/energy-pkg/. To deserialize into an exploitable numpy array, please use the function mendevi.database.serialize.binary_to_list().

Important

The perf library must be installed. Refer the perf guide.

Note

To access all information, the programme may require root privileges. Or you must give the permission with sudo sysctl -w kernel.perf_event_paranoid=0, otherwise, the sensor is not reachable. To make it permanent, create the file /etc/sysctl.d/99-perf.conf with the content kernel.perf_event_paranoid = 0.

act_wattmeter_dt¶

The list of time intervals in seconds between two consecutive wattmeter measures. To deserialize into an exploitable numpy array, please use the function mendevi.database.serialize.binary_to_list().

Tip

Only available for the measures taken on grid’5000.

act_wattmeter_power¶

The sampled external wattmeter measurement in w. To deserialize into an exploitable numpy array, please use the function mendevi.database.serialize.binary_to_list().

Tip

Only available for the measures taken on grid’5000.

act_gpu_dt¶

The list of time intervals in seconds between two consecutive GPU measures. To deserialize into an exploitable numpy array, please use the function mendevi.database.serialize.binary_to_list().

Tip

Only available for the measures taken with NVIDIA GPU.

act_gpu_power¶

The tensor of the sampled powers for each GPU. To deserialize into an exploitable numpy array, please use the function mendevi.database.serialize.binary_to_tensor().

Tip

Only available for the measures taken with NVIDIA GPU.

act_ps_dt¶

The list of the duration of each measurement point (in second). To deserialize into an exploitable numpy array, please use the function mendevi.database.serialize.binary_to_list().

act_ps_core¶

The percent usage of each logical core average activity for each interval. To deserialize into an exploitable numpy array, please use the function mendevi.database.serialize.binary_to_tensor().

act_ps_temp¶

The temperatures measured at each moment for all sensors. The data is presented in the form of a JSON dictionary that associates each sensor name with the list of temperatures recorded at each time interval act_ps_dt.

act_ps_ram¶

The sampled total usage of ram in bytes. To deserialize into an exploitable numpy array, please use the function mendevi.database.serialize.binary_to_list().

Table t_vid_video¶

vid_id	BINARY(128)	the video key
vid_name	TEXT	the basename of the video file
vid_codec	TINYTEXT	the small codec name
vid_duration	FLOAT	the video duration
vid_eotf	TINYTEXT	the color transfer name
vid_fps	FLOAT	the framerate
vid_frames	JSON	detailed content of all frames
vid_gamut	TINYTEXT	the color primaries name
vid_height	SMALLINT	video height in pixels
vid_pix_fmt	TINYTEXT	bit depth and subsampling name
vid_range	TINYTEXT	full range (pc) or limited (tv)
vid_size	BIGINT	file size in bytes
vid_width	SMALLINT	video width in pixels
vid_rms_sobel	LONGLONGBLOB	UVQ metric values
vid_rms_time_diff	LONGLONGBLOB	UVQ metric values
vid_uvq	LONGLONGBLOB	UVQ metric values

Contains the essential absolute characteristics of a video file.

vid_id¶

The md5 128 bits checksum of the video file. The names of transcoded videos include this signature encoded in base32.

vid_name¶

The base name of the video file, including the suffix:

sample_74d47vez2trcs3uupnjzclz4mm.mp4

vid_codec¶

The name of the video codec, catched by cutcutcodec.get_codec_video():

av1

vid_duration¶

The accurate video duration in second, including the display time of the last frame. Catched by cutcutcodec.get_duration_video().

vid_eotf¶

The name of the Electro Optical Transfer Function present in the metadata, or the most probable one if not. Catched by cutcutcodec.get_colorspace().transfer.

bt1361e, bt1361

vid_fps¶

The theorical framerate in Hz written in the file metadata. Catched by cutcutcodec.get_rate_video(). Very often, the exact value is a fraction with a denominator limited to 1001.

vid_frames¶

The list containing all the metadata for each frame.

[
    {
        "best_effort_timestamp": "0",
        "best_effort_timestamp_time": "0.000000",
        "chroma_location": "unspecified",
        "color_primaries": "smpte170m",
        "color_range": "tv",
        "color_space": "smpte170m",
        "color_transfer": "bt709",
        "crop_bottom": "0",
        "crop_left": "0",
        "crop_right": "0",
        "crop_top": "0",
        "duration": "1001",
        "duration_time": "0.041708",
        "height": "480",
        "interlaced_frame": "0",
        "key_frame": "1",
        "lossless": "0",
        "media_type": "video",
        "pict_type": "I",
        "pix_fmt": "yuv420p",
        "pkt_dts": "0",
        "pkt_dts_time": "0.000000",
        "pkt_pos": "48",
        "pkt_size": "147504",
        "pts": "0",
        "pts_time": "0.000000",
        "repeat_pict": "0",
        "sample_aspect_ratio": "1:1",
        "stream_index": "0",
        "top_field_first": "0",
        "width": "720"
    },
    ...
]

vid_gamut¶

The name of the colorspace primaries present in the metadata, or the most probable one if not. Catched by cutcutcodec.get_colorspace().primaries.

ntsc

vid_height¶

The height of the video in pixels, after being decoded.

vid_pix_fmt¶

The canonical name of the pixel format, carrying information about the colour space used (yuv or rgb), subsampling (444, 420 or 422) and bit depth (8, 10 or 12). Catched by cutcutcodec.get_pix_fmt().

yuv420p

vid_range¶

The encoded range as a string, pc for the full range, or tv for the limited range.

vid_size¶

The total size in bytes of the video file as it appears on the disc.

vid_width¶

The width of the video in pixels, after being decoded.

vid_rms_sobel¶

The serialised list of the spacial root means square sobel complexity. Catched by cutcutcodec.video_metrics(video, rms_sobel=True)["rms_sobel"]. To deserialize into an exploitable numpy array, please use the function mendevi.database.serialize.binary_to_list().

vid_rms_time_diff¶

The serialised list of the temporal root means square time difference. Catched by cutcutcodec.video_metrics(video, rms_time_diff=True)["rms_time_diff"]. To deserialize into an exploitable numpy array, please use the function mendevi.database.serialize.binary_to_list().

vid_uvq¶

The serialised list of the non comparative Google Universale Video Quality Model metric. Catched by cutcutcodec.video_metrics(video, uvq=True)["uvq"]. To deserialize into an exploitable numpy array, please use the function mendevi.database.serialize.binary_to_list().

Table t_met_metric¶

met_id	INTEGER	key
met_ref_vid_id	BINARY(128)	the reference video identifier
met_dis_vid_id	BINARY(128)	the identifier of the distorted video
met_lpips_alex	LONGBLOB	values of LPIPS ALEX metric
met_lpips_vgg	LONGBLOB	values of LPIPS VGG metric
met_psnr	LONGBLOB	values of PSNR metric
met_ssim	LONGBLOB	values of SSIM metric
met_vif	LONGBLOB	values of VIF metric
met_vmaf	LONGBLOB	values of VMAF metric

Compare the quality of a distorted video against a reference video using comparative metrics.

met_id¶

Unique primary key, with no particular meaning.

met_ref_vid_id¶

The never-null identifier of the reference video, as defined in vid_id.

met_dis_vid_id¶

The never-null identifier of the distorted video, as defined in vid_id.

met_lpips_alex¶

The serialised list of the LPIPS comparative metric using the ALEX neural network. Catched by cutcutcodec.video_metrics(dis, ref, lpips_alex=True)["lpips_alex"]. To deserialize into an exploitable numpy array, please use the function mendevi.database.serialize.binary_to_list().

met_lpips_vgg¶

The serialised list of the LPIPS comparative metric using the VGG neural network. Catched by cutcutcodec.video_metrics(dis, ref, lpips_vgg=True)["lpips_vgg"]. To deserialize into an exploitable numpy array, please use the function mendevi.database.serialize.binary_to_list().

met_psnr¶

The serialised list of the yuv (6, 1, 1) peak signal to noise ratio comparative metric. Catched by cutcutcodec.video_metrics(dis, ref, psnr=True)["psnr"]. To deserialize into an exploitable numpy array, please use the function mendevi.database.serialize.binary_to_list().

met_ssim¶

The serialised list of the yuv (6, 1, 1) (gauss win 11x11) structural similarity index measure comparative metric. Catched by cutcutcodec.video_metrics(dis, ref, ssim=True)["ssim"]. To deserialize into an exploitable numpy array, please use the function mendevi.database.serialize.binary_to_list().

met_vif¶

The serialised list of the visual information fidelity comparative metric on luminance. Catched by cutcutcodec.video_metrics(dis, ref, vif=True)["vif"]. To deserialize into an exploitable numpy array, please use the function mendevi.database.serialize.binary_to_list().

met_vmaf¶

The serialised list of the comparative metric Netflix Video Multi-Method Assessment Fusion metric. Catched by cutcutcodec.video_metrics(dis, ref, vmaf=True)["vmaf"]. To deserialize into an exploitable numpy array, please use the function mendevi.database.serialize.binary_to_list().

Table t_env_environment¶

env_id	INTEGER	Key
env_ffmpeg_version	MEDIUMTEXT	the full ffmpeg version
env_hostname	TINYTEXT	the username
env_kernel_version	TINYTEXT	the linux kernel version
env_librav1e_version	MEDIUMTEXT	the version of librav1e
env_libsvtav1_version	MEDIUMTEXT	the version of libsvtav1
env_libvpx_vp9_version	MEDIUMTEXT	the version of libvpx-vp9
env_libx265_version	MEDIUMTEXT	the version of libx265
env_logical_cores	INTEGER	the number of visible cpu
env_lshw	JSON	the machine config (lshw)
env_lspci	LONGTEXT	the machine config (lspci)
env_physical_cores	INTEGER	the real number of cpu
env_pip_freeze	MEDIUMTEXT	all python packages version
env_processor	TINYTEXT	the processos familly
env_python_compiler	TINYTEXT	name of python compiler
env_python_version	TINYTEXT	the python version
env_ram	INTEGER	Total amount of RAM
env_swap	INTEGER	Total amount of SWAP
env_system_version	MEDIUMTEXT	the OS version
env_vvc_version	MEDIUMTEXT	the version of vvenc
env_idle_act_id	INTEGER	IDLE as Table t_act_activity

Captures the physical characteristics and configuration of a machine. This table also captures the library version of a given measurement.

env_id¶

Unique primary key, with no particular meaning.

env_ffmpeg_version¶

The ffmpeg banner containing compilation options and library versions. This is human-readable text, returned by the function mendevi.measures.context.get_ffmpeg_version().

Please follow the :ref`install_ffmpeg`` section.

ffmpeg version N-120220-g934e1c23b0 Copyright (c) 2000-2025 the FFmpeg developers
built with gcc 14 (Ubuntu 14.2.0-4ubuntu2)
configuration: --prefix=/home/rrichard/ffmpeg_build --pkg-config-flags=--static --extra-cflags=-I/home/rrichard/ffmpeg_build/include --extra-ldflags=-L/home/rrichard/ffmpeg_build/lib --extra-libs='-lpthread -lm' --ld=g++ --bindir=/home/rrichard/bin --enable-gpl --enable-gnutls --enable-libaom --enable-libass --enable-libfdk-aac --enable-libfreetype --enable-libmp3lame --enable-libopus --enable-libsvtav1 --enable-libdav1d --enable-libvorbis --enable-libvpx --enable-libx264 --enable-libx265 --enable-libvvenc --enable-nonfree
libavutil      60.  4.101 / 60.  4.101
libavcodec     62.  6.100 / 62.  6.100
libavformat    62.  1.101 / 62.  1.101
libavdevice    62.  0.100 / 62.  0.100
libavfilter    11.  1.100 / 11.  1.100
libswscale      9.  0.100 /  9.  0.100
libswresample   6.  0.100 /  6.  0.100

env_hostname¶

The machine name, given by platform.node(), from the function mendevi.measures.context.get_platform().

paradoxe-32.rennes.grid5000.fr

env_kernel_version¶

The version of the Linux kernel, given by platform.platform(), from the function mendevi.measures.context.get_platform().

Linux-6.11.0-29-generic-x86_64-with-glibc2.40

env_librav1e_version¶

The version of rav1e excecutable file, whitch is a good candidate for ffmpeg librav1e encoder, from the function mendevi.measures.context.get_librav1e_version().

rav1e 0.8.0 (p20250624-1-gb7bf390) (release)
rustc 1.91.1 (ed61e7d7e 2025-11-07) x86_64-unknown-linux-gnu
Compiled CPU Features: adx,aes,avx,avx2,bmi1,bmi2,cmpxchg16b,f16c,fma,fxsr,lzcnt,movbe,pclmulqdq,popcnt,rdrand,rdseed,sse,sse2,sse3,sse4.1,sse4.2,ssse3,xsave,xsavec,xsaveopt,xsaves
Runtime Assembly Support: Enabled
Runtime Assembly Level: AVX2
Threading: Enabled
Unstable Features: Disabled
Compiler Flags: -C target-cpu=native

env_libsvtav1_version¶

The version of the libsvtav1 encoder and the version of the compiler for this library. It comes from the function mendevi.measures.context.get_libsvtav1_version().

v3.0.2-103-gc10214fa
GCC 14.2.0 64 bit

env_libvpx_vp9_version¶

The version of the libvpx-vp9 encoder. It comes from the function mendevi.measures.context.get_libvpx_vp9_version().

v1.14.1

env_libx265_version¶

The version of the libx265 encoder and the version of the compiler for this library. It comes from the function mendevi.measures.context.get_libx265_version().

4.1+188-cd4f0d6e9
[Linux][GCC 14.2.0][64 bit] 8bit+10bit+12bit
MMX2 SSE2Fast LZCNT SSSE3 SSE4.2 AVX FMA3 BMI2 AVX2

env_logical_cores¶

The number of logical (threaded) cores, in general more than the number of physical cores. It is given by psutil.cpu_count(logical=True), from the function mendevi.measures.context.get_platform().

env_lshw¶

A wealth of information on machine components, processor architecture, etc. Based on lshw -json.

Important

The lshw library must be installed. Refer the lshw guide.

Note

To access all information, the programme must have root privileges. Otherwise, some of the information will be missing.

{
    "id": "ptb-5cg4414jjd",
    "class": "system",
    "claimed": true,
    "description": "Computer",
    "width": 64,
    "capabilities": {
        "smp" : "Symmetric Multi-Processing",
        "vsyscall32" : "processus 32 bits"
    },
    "children" : [
        {
            "id": "core",
            "class": "bus",
            "claimed": true,
            "description": "Motherboard",
            "physid": "0",
            "children": ...
        },
        {
            "id" : "cpu",
            "class" : "processor",
            ...
            "configuration" : {
                "microcode" : "36"
            },
            "capabilities" : {
                "fpu" : "mathematical co-processor",
                ...
                "cpufreq" : "CPU Frequency scaling"
            }
        }
    ]
}

env_lspci¶

Informations on periferics, based on lspci -k.

env_physical_cores¶

The number of physical cores, in general less than the number of logical physical cores. It is given by psutil.cpu_count(logical=False), from the function mendevi.measures.context.get_platform().

env_pip_freeze¶

The list of the version of the python packages. The function mendevi.measures.context.get_pip_freeze() parse the command uv pip freeze.

aiohappyeyeballs==2.6.1
aiohttp==3.11.16
aiosignal==1.3.2
ansible-core==2.17.11
astroid==3.3.10
asttokens==3.0.0
attrs==25.3.0
av==14.3.0
bcrypt==4.3.0
black==25.1.0
cairocffi==1.7.1
CairoSVG==2.7.1
certifi==2025.1.31
cffi==1.17.1
charset-normalizer==3.4.1
click==8.1.8
colorama==0.4.6
...

env_processor¶

The proccessor familly, given by platform.processor(), from the function mendevi.measures.context.get_platform().

x86_64

env_python_compiler¶

The cython compiler and version, given by platform.python_compiler(), from the function mendevi.measures.context.get_platform().

GCC 14.2.0

env_python_version¶

The version of python, given by platform.python_version(), from the function mendevi.measures.context.get_platform().

3.13.2

env_ram¶

The total physical memory (exclusive swap) in bytes, given by psutil.virtual_memory().total, from the function mendevi.measures.context.get_platform().

env_swap¶

The total swap memory in bytes, given by psutil.swap_memory().total, from the function mendevi.measures.context.get_platform().

env_system_version¶

The name and the version of linux, given by platform.version(),from the function mendevi.measures.context.get_platform().

#29-Ubuntu SMP PREEMPT_DYNAMIC Fri Jun 13 20:29:41 UTC 2025

env_vvc_version¶

The version of the vvenc encoder. It comes from the function mendevi.measures.context.get_vvc_version().

1.13.1

env_idle_act_id¶

The id in the table Table t_act_activity of the IDLE measure of the machine.

Table t_dec_decode¶

dec_id	INTEGER	Key
dec_vid_id	BINARY(128)	the reference video identifier
dec_env_id	INTEGER	the environment identifier
dec_act_id	INTEGER	the activity identifier
dec_cmd	TEXT	ffmpeg command used
dec_decoder	TINYTEXT	the decoder name
dec_filter	TEXT	additional video filter
dec_height	SMALLINT	decoded height
dec_pix_fmt	TINYTEXT	decoded pixel format
dec_width	SMALLINT	decoded width

Contains the informations about video decoding.

dec_id¶

Unique primary key, with no particular meaning.

dec_vid_id¶

This is the reference of the source video that has been decoded. It points to the table Table t_vid_video.

dec_env_id¶

This is the reference of the context in which the decoding took place. It points to the table Table t_env_environment.

dec_act_id¶

This is a reference of measuring the machine’s activity during video decoding. It points to the table Table t_act_activity.

dec_cmd¶

The ffmpeg command used for decoding:

ffmpeg -threads 1 -i vid.mp4 -vf format=rgb24 -f null /dev/null

dec_decoder¶

The name of the decoder provided to ffmpeg. If nothing has been specified, this field takes the value null.

dec_filter¶

The additional ffmpeg video filter (after the -vf command) applied during decoding.

dec_height¶

In order to adapt the size of the decoded images to the size of the terminal, the frames are resized on the fly. Here is the final height of the screen.

dec_pix_fmt¶

During decoding, the pixels are converted into another format so that they can be displayed. This is the final format.

rgb24

dec_width¶

In order to adapt the size of the decoded images to the size of the terminal, the frames are resized on the fly. Here is the final width of the screen.

Table t_enc_encode¶

enc_id	INTEGER	Key
enc_src_vid_id	BINARY(128)	the input video
enc_dst_vid_id	BINARY(128)	the ouput transcoded video
enc_env_id	INTEGER	the environment identifier
enc_act_id	INTEGER	the activity identifier
enc_cmd	TEXT	ffmpeg command used
enc_effort	TINYTEXT	the effort made by the encoder
enc_encoder	TINYTEXT	the encoder name
enc_filter	TEXT	additional video filter
enc_fps	FLOAT	the input framerate
enc_height	SMALLINT	the input frames height
env_mode	TINYTEXT	constant or variable bitrate
enc_pix_fmt	TINYTEXT	the input pixel format
enc_quality	FLOAT	the encoding crf or qp
enc_threads	SMALLINT	the number of threads used
enc_width	SMALLINT	the input frames width

Contains the informations about video encoding.

enc_id¶

Unique primary key, with no particular meaning.

enc_src_vid_id¶

This is the reference of the input video that enters into the encoder. It points to the table Table t_vid_video.

enc_dst_vid_id¶

This is the reference of the transcoded video that exits the encoder. It points to the table Table t_vid_video.

enc_env_id¶

This is the reference of the context in which the encoding took place. It points to the table Table t_env_environment.

enc_act_id¶

This is a reference of measuring the machine’s activity during video encoding. It points to the table Table t_act_activity.

enc_cmd¶

The ffmpeg command used for encoding:

ffmpeg -y -i ref.mp4 -c:v libsvtav1 -crf 15 -preset 6 -svtav1-params film-grain=0:lp=3 dst.mp4

enc_effort¶

The effort that the encoder puts into optimizing the bit rate distortion ratio in exchange for increased computing time. Possible values are fast, medium and slow.

enc_encoder¶

The ffmpeg encoder name. Possible values are libx264, libx265, libvpx-vp9, libsvtav1 and vvc.

enc_filter¶

The additional ffmpeg video filter (after the -vf command) applied during decoding.

enc_fps¶

The frame rate in Hz of the video at the encoder input.

enc_height¶

The number of pixel of the frames along the height dimension.

enc_mode¶

Text that is cbr if the bit rate is constant (cbr mode), or vbr if it is variable (constant quality mode).

enc_pix_fmt¶

The pixel format used for encoding.

yuv420p

enc_quality¶

The relative quality \(\in [0, 1]\) that the encoder aims for the final video. This number is afinally related to the crf or qp parameter. 0 corresponds to near lossless encoding, and 1 to highly lossly encoding.

enc_threads¶

The number of threads that the encoder uses simultaneously. This is related to the CPU utilisation rate.

enc_width¶

The number of pixel of the frames along the width dimension.