max |direct - evidently| over 80 checks: 0.000000

no shift, R=5000, reference = everything else
batch           frames  dataset  timeofday    weather   vehicles       sign  mean d wx  mean d tod
frames 1470       1470     0.0%       0.0%       0.0%       0.0%       0.0%      0.015       0.012
videos 10         1471    57.3%      88.3%      94.8%      57.2%       4.7%      0.239       0.189
frames 2940       2940     0.0%       0.0%       0.0%       0.0%       0.0%      0.011       0.009
videos 20         2940    15.0%      61.6%      76.2%      29.3%       0.6%      0.167       0.131
frames 5880       5880     0.0%       0.0%       0.0%       0.0%       0.0%      0.008       0.007
videos 40         5878     2.0%      37.8%      62.6%      11.0%       0.0%      0.119       0.092
videos 80        11763     0.1%      25.0%      43.3%       4.5%       0.0%      0.097       0.075

batch of contiguous videos in recording order vs shuffled videos, reference = everything else, R=5000
 videos  wx shuffled  wx contiguous  tod shuffled  tod contiguous
     10        0.242          0.380         0.189           0.288
     20        0.165          0.365         0.130           0.276
     40        0.116          0.370         0.092           0.266
neighbouring videos with the same value: {'timeofday': 0.94, 'weather': 0.88} runs: {'timeofday': 11, 'weather': 21}

recording order: thresholds resampled inside the first 60 videos, batches of 10 after them
{'timeofday': 0.133, 'weather': 0.226, 'traffic': 0.267, 'vehicles': 0.204, 'sign': 0.109, 'traffic_light': 0.147, 'pedestrian': 0.088, 'opposite_lane': 0.199}
videos     frames  over default  over resampled  columns over resampled
061-070      1184             4               4  traffic vehicles sign opposite_lane
071-081      1560             6               4  vehicles sign traffic_light opposite_lane
082-091      1320             3               2  vehicles opposite_lane
092-101       762             5               1  weather
102-111      1800             4               3  weather vehicles sign
112-121      1800             6               5  weather vehicles sign traffic_light opposite_lane
122-131      1800             5               4  weather traffic vehicles sign
132-141      1800             5               2  vehicles opposite_lane
142-151      1800             7               4  timeofday vehicles traffic_light opposite_lane
152-161      1653             6               5  timeofday weather vehicles sign opposite_lane
162-171      1697             4               3  timeofday weather vehicles
172-172        95             7               7  timeofday weather traffic vehicles sign traffic_light opposite_lane

dark videos fixed in the batch, reference = half of the videos, batch of 20 from the other half, R=5000
dark videos 24 of 171
dark in batch   dark frames  timeofday  dataset
random                  15%      59.3%    17.0%
3/20                    16%      35.4%    11.7%
6/20                    32%      98.0%    26.1%
10/20                    53%     100.0%    38.9%
12/20                    63%     100.0%    47.8%