Gnuplot not found, using plotters backend
Benchmarking grayscale_u8_rgb/256
Benchmarking grayscale_u8_rgb/256: Warming up for 500.00 ms
Benchmarking grayscale_u8_rgb/256: Collecting 20 samples in estimated 2.0124 s (28k iterations)
Benchmarking grayscale_u8_rgb/256: Analyzing
grayscale_u8_rgb/256    time:   [18.577 µs 23.364 µs 27.290 µs]
                        change: [-15.455% -1.0107% +17.392%] (p = 0.91 > 0.05)
                        No change in performance detected.
Found 5 outliers among 20 measurements (25.00%)
  4 (20.00%) high mild
  1 (5.00%) high severe
Benchmarking grayscale_u8_rgb/512
Benchmarking grayscale_u8_rgb/512: Warming up for 500.00 ms
Benchmarking grayscale_u8_rgb/512: Collecting 20 samples in estimated 2.0196 s (7350 iterations)
Benchmarking grayscale_u8_rgb/512: Analyzing
grayscale_u8_rgb/512    time:   [64.188 µs 65.272 µs 66.820 µs]
                        change: [-6.1332% -1.7066% +2.4074%] (p = 0.47 > 0.05)
                        No change in performance detected.
Found 3 outliers among 20 measurements (15.00%)
  3 (15.00%) high mild
Benchmarking grayscale_u8_rgb/1024
Benchmarking grayscale_u8_rgb/1024: Warming up for 500.00 ms
Benchmarking grayscale_u8_rgb/1024: Collecting 20 samples in estimated 2.0167 s (3570 iterations)
Benchmarking grayscale_u8_rgb/1024: Analyzing
grayscale_u8_rgb/1024   time:   [262.85 µs 268.76 µs 277.44 µs]
                        change: [-11.146% -7.1871% -3.0655%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 20 measurements (10.00%)
  2 (10.00%) high mild

Benchmarking grayscale_f32_rgb/256
Benchmarking grayscale_f32_rgb/256: Warming up for 500.00 ms
Benchmarking grayscale_f32_rgb/256: Collecting 20 samples in estimated 2.0163 s (14k iterations)
Benchmarking grayscale_f32_rgb/256: Analyzing
grayscale_f32_rgb/256   time:   [94.478 µs 96.955 µs 99.764 µs]
                        change: [-9.5725% -5.0515% -0.3334%] (p = 0.04 < 0.05)
                        Change within noise threshold.
Found 3 outliers among 20 measurements (15.00%)
  1 (5.00%) low mild
  2 (10.00%) high severe
Benchmarking grayscale_f32_rgb/512
Benchmarking grayscale_f32_rgb/512: Warming up for 500.00 ms
Benchmarking grayscale_f32_rgb/512: Collecting 20 samples in estimated 2.0368 s (2940 iterations)
Benchmarking grayscale_f32_rgb/512: Analyzing
grayscale_f32_rgb/512   time:   [405.30 µs 415.40 µs 430.76 µs]
                        change: [-5.1957% -1.3093% +2.6809%] (p = 0.53 > 0.05)
                        No change in performance detected.
Found 4 outliers among 20 measurements (20.00%)
  4 (20.00%) high mild
Benchmarking grayscale_f32_rgb/1024
Benchmarking grayscale_f32_rgb/1024: Warming up for 500.00 ms
Benchmarking grayscale_f32_rgb/1024: Collecting 20 samples in estimated 2.3261 s (420 iterations)
Benchmarking grayscale_f32_rgb/1024: Analyzing
grayscale_f32_rgb/1024  time:   [2.3110 ms 2.4322 ms 2.5984 ms]
                        change: [-12.937% -4.9976% +3.6501%] (p = 0.27 > 0.05)
                        No change in performance detected.
Found 1 outliers among 20 measurements (5.00%)
  1 (5.00%) high mild

Benchmarking threshold_u8/256
Benchmarking threshold_u8/256: Warming up for 500.00 ms
Benchmarking threshold_u8/256: Collecting 20 samples in estimated 2.0006 s (299k iterations)
Benchmarking threshold_u8/256: Analyzing
threshold_u8/256        time:   [3.0871 µs 3.3087 µs 3.4545 µs]
                        change: [-7.7183% +6.0960% +22.852%] (p = 0.42 > 0.05)
                        No change in performance detected.
Benchmarking threshold_u8/512
Benchmarking threshold_u8/512: Warming up for 500.00 ms
Benchmarking threshold_u8/512: Collecting 20 samples in estimated 2.0017 s (84k iterations)
Benchmarking threshold_u8/512: Analyzing
threshold_u8/512        time:   [10.708 µs 11.376 µs 11.816 µs]
                        change: [-3.3752% +8.7812% +22.149%] (p = 0.18 > 0.05)
                        No change in performance detected.
Benchmarking threshold_u8/1024
Benchmarking threshold_u8/1024: Warming up for 500.00 ms
Benchmarking threshold_u8/1024: Collecting 20 samples in estimated 2.0262 s (16k iterations)
Benchmarking threshold_u8/1024: Analyzing
threshold_u8/1024       time:   [44.609 µs 46.905 µs 49.208 µs]
                        change: [+0.2292% +5.3826% +10.893%] (p = 0.06 > 0.05)
                        No change in performance detected.
Found 1 outliers among 20 measurements (5.00%)
  1 (5.00%) high mild

Benchmarking threshold_u8_cropped/256
Benchmarking threshold_u8_cropped/256: Warming up for 500.00 ms
Benchmarking threshold_u8_cropped/256: Collecting 20 samples in estimated 2.0013 s (202k iterations)
Benchmarking threshold_u8_cropped/256: Analyzing
threshold_u8_cropped/256
                        time:   [7.1822 µs 7.4604 µs 7.8206 µs]
                        change: [-4.8093% +0.9598% +7.5211%] (p = 0.78 > 0.05)
                        No change in performance detected.
Found 2 outliers among 20 measurements (10.00%)
  1 (5.00%) high mild
  1 (5.00%) high severe
Benchmarking threshold_u8_cropped/512
Benchmarking threshold_u8_cropped/512: Warming up for 500.00 ms
Benchmarking threshold_u8_cropped/512: Collecting 20 samples in estimated 2.0064 s (66k iterations)
Benchmarking threshold_u8_cropped/512: Analyzing
threshold_u8_cropped/512
                        time:   [18.341 µs 19.471 µs 20.664 µs]
                        change: [-1.3267% +6.3580% +13.563%] (p = 0.13 > 0.05)
                        No change in performance detected.
Benchmarking threshold_u8_cropped/1024
Benchmarking threshold_u8_cropped/1024: Warming up for 500.00 ms
Benchmarking threshold_u8_cropped/1024: Collecting 20 samples in estimated 2.0286 s (13k iterations)
Benchmarking threshold_u8_cropped/1024: Analyzing
threshold_u8_cropped/1024
                        time:   [92.121 µs 93.382 µs 94.566 µs]
                        change: [-1.3435% +0.1634% +1.7021%] (p = 0.84 > 0.05)
                        No change in performance detected.
Found 1 outliers among 20 measurements (5.00%)
  1 (5.00%) high mild

Benchmarking cast_f32_to_u8/256
Benchmarking cast_f32_to_u8/256: Warming up for 500.00 ms
Benchmarking cast_f32_to_u8/256: Collecting 20 samples in estimated 2.0118 s (10k iterations)
Benchmarking cast_f32_to_u8/256: Analyzing
cast_f32_to_u8/256      time:   [133.94 µs 137.26 µs 142.87 µs]
                        change: [-8.2142% -4.5225% -0.9074%] (p = 0.03 < 0.05)
                        Change within noise threshold.
Found 2 outliers among 20 measurements (10.00%)
  1 (5.00%) high mild
  1 (5.00%) high severe
Benchmarking cast_f32_to_u8/512
Benchmarking cast_f32_to_u8/512: Warming up for 500.00 ms
Benchmarking cast_f32_to_u8/512: Collecting 20 samples in estimated 2.1755 s (2520 iterations)
Benchmarking cast_f32_to_u8/512: Analyzing
cast_f32_to_u8/512      time:   [550.43 µs 560.24 µs 569.49 µs]
                        change: [-6.2565% -2.3007% +0.7091%] (p = 0.27 > 0.05)
                        No change in performance detected.
Found 2 outliers among 20 measurements (10.00%)
  1 (5.00%) high mild
  1 (5.00%) high severe
Benchmarking cast_f32_to_u8/1024
Benchmarking cast_f32_to_u8/1024: Warming up for 500.00 ms
Benchmarking cast_f32_to_u8/1024: Collecting 20 samples in estimated 2.6029 s (420 iterations)
Benchmarking cast_f32_to_u8/1024: Analyzing
cast_f32_to_u8/1024     time:   [2.7517 ms 2.8490 ms 2.9539 ms]
                        change: [+5.7751% +10.387% +15.263%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 1 outliers among 20 measurements (5.00%)
  1 (5.00%) high severe

Benchmarking cast_u8_to_f32/256
Benchmarking cast_u8_to_f32/256: Warming up for 500.00 ms
Benchmarking cast_u8_to_f32/256: Collecting 20 samples in estimated 2.0068 s (34k iterations)
Benchmarking cast_u8_to_f32/256: Analyzing
cast_u8_to_f32/256      time:   [43.452 µs 44.706 µs 45.693 µs]
                        change: [-8.6149% -1.0040% +7.3392%] (p = 0.82 > 0.05)
                        No change in performance detected.
Benchmarking cast_u8_to_f32/512
Benchmarking cast_u8_to_f32/512: Warming up for 500.00 ms
Benchmarking cast_u8_to_f32/512: Collecting 20 samples in estimated 2.0308 s (8190 iterations)
Benchmarking cast_u8_to_f32/512: Analyzing
cast_u8_to_f32/512      time:   [171.36 µs 173.01 µs 176.34 µs]
                        change: [-4.9110% -2.3720% +0.3750%] (p = 0.10 > 0.05)
                        No change in performance detected.
Found 2 outliers among 20 measurements (10.00%)
  1 (5.00%) high mild
  1 (5.00%) high severe
Benchmarking cast_u8_to_f32/1024
Benchmarking cast_u8_to_f32/1024: Warming up for 500.00 ms
Benchmarking cast_u8_to_f32/1024: Collecting 20 samples in estimated 2.1900 s (1260 iterations)
Benchmarking cast_u8_to_f32/1024: Analyzing
cast_u8_to_f32/1024     time:   [1.0958 ms 1.1323 ms 1.1728 ms]
                        change: [-18.534% -13.115% -6.8055%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 20 measurements (5.00%)
  1 (5.00%) high severe

Benchmarking materialize_flip_h_u8/256
Benchmarking materialize_flip_h_u8/256: Warming up for 500.00 ms
Benchmarking materialize_flip_h_u8/256: Collecting 20 samples in estimated 2.0074 s (33k iterations)
Benchmarking materialize_flip_h_u8/256: Analyzing
materialize_flip_h_u8/256
                        time:   [53.570 µs 55.740 µs 58.759 µs]
                        change: [-85.312% -84.377% -83.561%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 20 measurements (10.00%)
  1 (5.00%) high mild
  1 (5.00%) high severe
Benchmarking materialize_flip_h_u8/512
Benchmarking materialize_flip_h_u8/512: Warming up for 500.00 ms
Benchmarking materialize_flip_h_u8/512: Collecting 20 samples in estimated 2.0101 s (6930 iterations)
Benchmarking materialize_flip_h_u8/512: Analyzing
materialize_flip_h_u8/512
                        time:   [225.88 µs 230.11 µs 233.93 µs]
                        change: [-84.933% -84.114% -83.345%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 20 measurements (5.00%)
  1 (5.00%) high mild
Benchmarking materialize_flip_h_u8/1024
Benchmarking materialize_flip_h_u8/1024: Warming up for 500.00 ms
Benchmarking materialize_flip_h_u8/1024: Collecting 20 samples in estimated 2.2848 s (1680 iterations)
Benchmarking materialize_flip_h_u8/1024: Analyzing
materialize_flip_h_u8/1024
                        time:   [1.0693 ms 1.1260 ms 1.1940 ms]
                        change: [-81.392% -80.481% -79.559%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 20 measurements (5.00%)
  1 (5.00%) high mild

Benchmarking materialize_flip_v_u8/256
Benchmarking materialize_flip_v_u8/256: Warming up for 500.00 ms
Benchmarking materialize_flip_v_u8/256: Collecting 20 samples in estimated 2.0016 s (100k iterations)
Benchmarking materialize_flip_v_u8/256: Analyzing
materialize_flip_v_u8/256
                        time:   [9.9041 µs 10.345 µs 10.640 µs]
                        change: [-97.562% -97.323% -97.095%] (p = 0.00 < 0.05)
                        Performance has improved.
Benchmarking materialize_flip_v_u8/512
Benchmarking materialize_flip_v_u8/512: Warming up for 500.00 ms
Benchmarking materialize_flip_v_u8/512: Collecting 20 samples in estimated 2.0034 s (19k iterations)
Benchmarking materialize_flip_v_u8/512: Analyzing
materialize_flip_v_u8/512
                        time:   [57.224 µs 57.877 µs 59.002 µs]
                        change: [-95.839% -95.708% -95.577%] (p = 0.00 < 0.05)
                        Performance has improved.
Benchmarking materialize_flip_v_u8/1024
Benchmarking materialize_flip_v_u8/1024: Warming up for 500.00 ms
Benchmarking materialize_flip_v_u8/1024: Collecting 20 samples in estimated 2.0591 s (3990 iterations)
Benchmarking materialize_flip_v_u8/1024: Analyzing
materialize_flip_v_u8/1024
                        time:   [271.85 µs 277.45 µs 283.43 µs]
                        change: [-95.486% -95.195% -94.961%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 20 measurements (15.00%)
  2 (10.00%) low mild
  1 (5.00%) high mild

Benchmarking materialize_transpose_u8/256
Benchmarking materialize_transpose_u8/256: Warming up for 500.00 ms
Benchmarking materialize_transpose_u8/256: Collecting 20 samples in estimated 2.0067 s (29k iterations)
Benchmarking materialize_transpose_u8/256: Analyzing
materialize_transpose_u8/256
                        time:   [59.162 µs 62.470 µs 66.166 µs]
                        change: [-83.755% -82.925% -82.044%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 20 measurements (15.00%)
  1 (5.00%) high mild
  2 (10.00%) high severe
Benchmarking materialize_transpose_u8/512
Benchmarking materialize_transpose_u8/512: Warming up for 500.00 ms
Benchmarking materialize_transpose_u8/512: Collecting 20 samples in estimated 2.0343 s (5670 iterations)
Benchmarking materialize_transpose_u8/512: Analyzing
materialize_transpose_u8/512
                        time:   [307.55 µs 313.45 µs 323.68 µs]
                        change: [-79.501% -77.977% -76.147%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 4 outliers among 20 measurements (20.00%)
  1 (5.00%) high mild
  3 (15.00%) high severe
Benchmarking materialize_transpose_u8/1024
Benchmarking materialize_transpose_u8/1024: Warming up for 500.00 ms
Benchmarking materialize_transpose_u8/1024: Collecting 20 samples in estimated 2.7295 s (630 iterations)
Benchmarking materialize_transpose_u8/1024: Analyzing
materialize_transpose_u8/1024
                        time:   [3.9760 ms 4.1479 ms 4.3879 ms]
                        change: [-72.657% -71.688% -70.677%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 20 measurements (5.00%)
  1 (5.00%) high severe

Benchmarking cast_u8_to_f32_flip_h/256
Benchmarking cast_u8_to_f32_flip_h/256: Warming up for 500.00 ms
Benchmarking cast_u8_to_f32_flip_h/256: Collecting 20 samples in estimated 2.0075 s (19k iterations)
Benchmarking cast_u8_to_f32_flip_h/256: Analyzing
cast_u8_to_f32_flip_h/256
                        time:   [89.405 µs 91.758 µs 93.932 µs]
                        change: [-80.270% -79.569% -78.953%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 20 measurements (5.00%)
  1 (5.00%) high mild
Benchmarking cast_u8_to_f32_flip_h/512
Benchmarking cast_u8_to_f32_flip_h/512: Warming up for 500.00 ms
Benchmarking cast_u8_to_f32_flip_h/512: Collecting 20 samples in estimated 2.0132 s (3990 iterations)
Benchmarking cast_u8_to_f32_flip_h/512: Analyzing
cast_u8_to_f32_flip_h/512
                        time:   [373.27 µs 391.19 µs 410.43 µs]
                        change: [-78.113% -77.310% -76.513%] (p = 0.00 < 0.05)
                        Performance has improved.
Benchmarking cast_u8_to_f32_flip_h/1024
Benchmarking cast_u8_to_f32_flip_h/1024: Warming up for 500.00 ms
Benchmarking cast_u8_to_f32_flip_h/1024: Collecting 20 samples in estimated 2.0994 s (630 iterations)
Benchmarking cast_u8_to_f32_flip_h/1024: Analyzing
cast_u8_to_f32_flip_h/1024
                        time:   [2.3707 ms 2.4371 ms 2.5083 ms]
                        change: [-71.698% -69.734% -68.131%] (p = 0.00 < 0.05)
                        Performance has improved.

Benchmarking scale_f32_transposed/256
Benchmarking scale_f32_transposed/256: Warming up for 500.00 ms
Benchmarking scale_f32_transposed/256: Collecting 20 samples in estimated 2.0126 s (7770 iterations)
Benchmarking scale_f32_transposed/256: Analyzing
scale_f32_transposed/256
                        time:   [186.94 µs 194.33 µs 205.25 µs]
                        change: [-84.858% -83.773% -82.643%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 4 outliers among 20 measurements (20.00%)
  1 (5.00%) high mild
  3 (15.00%) high severe
Benchmarking scale_f32_transposed/512
Benchmarking scale_f32_transposed/512: Warming up for 500.00 ms
Benchmarking scale_f32_transposed/512: Collecting 20 samples in estimated 2.0939 s (1050 iterations)
Benchmarking scale_f32_transposed/512: Analyzing
scale_f32_transposed/512
                        time:   [1.4013 ms 1.4386 ms 1.4934 ms]
                        change: [-80.311% -79.025% -77.684%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 20 measurements (5.00%)
  1 (5.00%) high mild
Benchmarking scale_f32_transposed/1024
Benchmarking scale_f32_transposed/1024: Warming up for 500.00 ms
Benchmarking scale_f32_transposed/1024: Collecting 20 samples in estimated 2.3560 s (80 iterations)
Benchmarking scale_f32_transposed/1024: Analyzing
scale_f32_transposed/1024
                        time:   [17.218 ms 17.535 ms 17.919 ms]
                        change: [-71.195% -69.874% -68.676%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 20 measurements (10.00%)
  1 (5.00%) high mild
  1 (5.00%) high severe

Benchmarking grayscale_u8_flip_h/256
Benchmarking grayscale_u8_flip_h/256: Warming up for 500.00 ms
Benchmarking grayscale_u8_flip_h/256: Collecting 20 samples in estimated 2.0123 s (24k iterations)
Benchmarking grayscale_u8_flip_h/256: Analyzing
grayscale_u8_flip_h/256 time:   [71.000 µs 72.034 µs 73.912 µs]
                        change: [-79.985% -79.614% -79.151%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 4 outliers among 20 measurements (20.00%)
  2 (10.00%) high mild
  2 (10.00%) high severe
Benchmarking grayscale_u8_flip_h/512
Benchmarking grayscale_u8_flip_h/512: Warming up for 500.00 ms
Benchmarking grayscale_u8_flip_h/512: Collecting 20 samples in estimated 2.0274 s (5880 iterations)
Benchmarking grayscale_u8_flip_h/512: Analyzing
grayscale_u8_flip_h/512 time:   [287.62 µs 291.23 µs 295.95 µs]
                        change: [-80.058% -79.720% -79.316%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 20 measurements (5.00%)
  1 (5.00%) high mild
Benchmarking grayscale_u8_flip_h/1024
Benchmarking grayscale_u8_flip_h/1024: Warming up for 500.00 ms
Benchmarking grayscale_u8_flip_h/1024: Collecting 20 samples in estimated 2.0427 s (1260 iterations)
Benchmarking grayscale_u8_flip_h/1024: Analyzing
grayscale_u8_flip_h/1024
                        time:   [1.3052 ms 1.3317 ms 1.3573 ms]
                        change: [-80.676% -79.159% -77.818%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 20 measurements (5.00%)
  1 (5.00%) high mild

Benchmarking resize_224_u8/256
Benchmarking resize_224_u8/256: Warming up for 500.00 ms
Benchmarking resize_224_u8/256: Collecting 20 samples in estimated 2.0185 s (6720 iterations)
Benchmarking resize_224_u8/256: Analyzing
resize_224_u8/256       time:   [261.14 µs 270.69 µs 285.77 µs]
                        change: [-2.0242% +3.1216% +10.124%] (p = 0.36 > 0.05)
                        No change in performance detected.
Found 3 outliers among 20 measurements (15.00%)
  3 (15.00%) high severe
Benchmarking resize_224_u8/512
Benchmarking resize_224_u8/512: Warming up for 500.00 ms
Benchmarking resize_224_u8/512: Collecting 20 samples in estimated 2.0149 s (4410 iterations)
Benchmarking resize_224_u8/512: Analyzing
resize_224_u8/512       time:   [393.35 µs 396.94 µs 401.23 µs]
                        change: [-19.747% -9.8036% -0.8006%] (p = 0.07 > 0.05)
                        No change in performance detected.
Found 3 outliers among 20 measurements (15.00%)
  3 (15.00%) high mild
Benchmarking resize_224_u8/1024
Benchmarking resize_224_u8/1024: Warming up for 500.00 ms
Benchmarking resize_224_u8/1024: Collecting 20 samples in estimated 2.2374 s (1680 iterations)
Benchmarking resize_224_u8/1024: Analyzing
resize_224_u8/1024      time:   [846.15 µs 856.45 µs 867.13 µs]
                        change: [-13.549% -5.7514% +0.0111%] (p = 0.15 > 0.05)
                        No change in performance detected.

Benchmarking crop_then_resize_224_u8/256
Benchmarking crop_then_resize_224_u8/256: Warming up for 500.00 ms
Benchmarking crop_then_resize_224_u8/256: Collecting 20 samples in estimated 2.0124 s (8190 iterations)
Benchmarking crop_then_resize_224_u8/256: Analyzing
crop_then_resize_224_u8/256
                        time:   [226.85 µs 236.53 µs 249.95 µs]
                        change: [-46.140% -44.187% -41.951%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 20 measurements (15.00%)
  3 (15.00%) high mild
Benchmarking crop_then_resize_224_u8/512
Benchmarking crop_then_resize_224_u8/512: Warming up for 500.00 ms
Benchmarking crop_then_resize_224_u8/512: Collecting 20 samples in estimated 2.0689 s (4200 iterations)
Benchmarking crop_then_resize_224_u8/512: Analyzing
crop_then_resize_224_u8/512
                        time:   [385.86 µs 392.37 µs 401.11 µs]
                        change: [-67.434% -65.940% -64.527%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 20 measurements (10.00%)
  2 (10.00%) high mild
Benchmarking crop_then_resize_224_u8/1024
Benchmarking crop_then_resize_224_u8/1024: Warming up for 500.00 ms
Benchmarking crop_then_resize_224_u8/1024: Collecting 20 samples in estimated 2.0158 s (1680 iterations)
Benchmarking crop_then_resize_224_u8/1024: Analyzing
crop_then_resize_224_u8/1024
                        time:   [810.27 µs 846.85 µs 895.05 µs]
                        change: [-79.655% -78.144% -76.520%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 20 measurements (5.00%)
  1 (5.00%) high mild
