ACES Shoot - Colour Checker Detection

Thomas_Mansencal · August 7, 2021, 9:59am

Hi,

Starting a thread to document the progress with the current work for detecting the colour checkers in the shoot @Giardiello did back in May with a bunch of cameras. Using a Google Spreadsheet, Google Colab, TIFF extractions of the relevant camera clips and the develop branch of Colour Checker Detection, we have now an export of the samples for each camera bracketing group:

gist.github.com

https://gist.github.com/KelSolaar/7f416f2da4d559611193c05c9776fbbe

aces_shoot_cc24_samples.json

{
    "ALEXA SXT": {
        "A-2_2": {
            "-2": {
                "Directory": "/20210601/A003R732/A003R732/A003C003_210519_R732",
                "Samples (Sequence)": [
                    [
                        [0.2347482293844223, 0.18081583082675934, 0.14596495032310486],
                        [0.3502565324306488, 0.2826744616031647, 0.2634207308292389],
                        [0.22467496991157532, 0.25057387351989746, 0.26863107085227966],

This file has been truncated. show original

Here are the two current Colab notebooks:

There are a few kinks to work out but there should be enough data to get going!

Cheers,

Thomas

priikone · August 7, 2021, 12:30pm

This is great @Thomas_Mansencal!

Is one of the colab missing? Both link to same notebook…

Thomas_Mansencal · August 7, 2021, 7:45pm

Thanks fixed! Copy-paste failure!

Thomas_Mansencal · August 10, 2021, 9:53am

Looking at a subset of the data, i.e. the Neutral 5 (.70D), I noted that some of the images had significantly higher exposure:

            "0": {
                "Directory": "/20210518/B001/B001_C034_20130511_P00000",
                "Encoding": "DJI-LOG/DJI-GAMUT",
                "Neutral 5 (.70 D) (Median)": [0.42103683948516846, 0.41377878189086914, 0.40679672360420227],
                "Neutral 5 (.70 D) (Sequence)": [
                    [0.4211159944534302, 0.41391751170158386, 0.40676257014274597],
                    [0.42103949189186096, 0.41377878189086914, 0.40671026706695557],
                    [0.42108091711997986, 0.4136505126953125, 0.40679672360420227],
                    [0.4211457669734955, 0.4136641323566437, 0.4067840874195099],
                    [0.42084261775016785, 0.4137943983078003, 0.406970351934433],
                    [0.42103683948516846, 0.4138261675834656, 0.4072221517562866],
                    [0.4209568500518799, 0.41409289836883545, 0.4071600139141083],
                    [0.4209418296813965, 0.4138721823692322, 0.4066964089870453],
                    [0.4211048185825348, 0.4137803018093109, 0.4072290062904358],
                    [0.4211867153644562, 0.4138403832912445, 0.4072200357913971],
                    [0.42087313532829285, 0.41367125511169434, 0.4067012667655945],
                    [0.4209837317466736, 0.41356655955314636, 0.4067242443561554],
                    [0.4206361770629883, 0.4136791527271271, 0.40718263387680054],
                    [0.42101505398750305, 0.41375190019607544, 0.4065482020378113],
                    [0.4210021495819092, 0.41375789046287537, 0.4064512252807617],
                    [0.42094552516937256, 0.4137341380119324, 0.4065594971179962],
                    [0.4209025502204895, 0.41344916820526123, 0.40667569637298584],
              --->  [0.5510446429252625, 0.541676938533783, 0.5352358222007751],  <---
                    [0.4210505187511444, 0.4141332805156708, 0.4072085916996002],
                    [0.4210974872112274, 0.4137188792228699, 0.4065997004508972],
                    [0.4211079180240631, 0.4139573574066162, 0.4072112739086151],
                    [0.4210573136806488, 0.41371357440948486, 0.407023161649704],
                    [0.42107313871383667, 0.4136336147785187, 0.40621843934059143],
                    [0.42101046442985535, 0.41382649540901184, 0.40690770745277405],
                    [0.42101871967315674, 0.4139617681503296, 0.40716755390167236],
                    [0.4211474359035492, 0.4142133295536041, 0.4069019556045532]
                ]
            }
        }
    },

With that in mind, I have rerun the processing notebook with outlier rejection to cull off the offenders and get a better median. I have also added the encoding as metadata doing so.

The Neutral 5 subset is here: ACES Shoot - CC24 Neutral 5 (.70 D) · GitHub

Cheers,

Thomas

Thomas_Mansencal · August 22, 2021, 4:05am

Yesterday, I updated the code to also extra the Great COD (Grey Card) values. The Gist data has been re-generated accordingly.

Thomas_Mansencal · September 7, 2021, 10:10am

Hello!

I updated/refactored the notebook yesterday so that it is easy to extract the flat field data while improving flipped checker management. The fields are actually really good, here are the median values for the colour checker 6x4 samples for some of the cameras:

And min-max normalised:

I have also implemented support for semi-automatic spreasheet generation: ACES IDT VWG - ACES Shoot - Analysis - Google Sheets

Cheers,

Thomas