[ { "source": "14", "label": "Railway", "selection_rule": "Preserved v13 example: largest MOS reversal per distinct source where lens agrees and VMAF reverses.", "clips": [ { "lane": "Ref", "source_file": "data/video/waterloo4k/ref/14.mp4", "source_sha256": "331ede7678276a66caf2d1cbf49193a9167dfeafefc10075f6f550d526bedb16", "source_stream": { "codec_name": "h264", "width": 3840, "height": 2160, "r_frame_rate": "30000/1001", "avg_frame_rate": "30000/1001", "time_base": "1/30000", "duration": "10.010000", "nb_frames": "300" }, "frame_indices_zero_based": [ 122, 130, 138 ], "display_canvas_pixels": [ 1920, 1080 ], "filter": "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "frames": [ { "source_frame_index": 122, "source_pts": 122122, "timestamp_seconds": 4.070733333333333, "nominal_timestamp_seconds": 4.070733333333333, "decoded_png_sha256": "399f7be541849281a4b6c7f153f0bb989a3a8acc122127769623e23a5081eb52", "rgb_array_sha256": "ac757868ebbe22b754a551f134be41bf7e3e39baedbb7700fd3c97d709e82e88" }, { "source_frame_index": 130, "source_pts": 130130, "timestamp_seconds": 4.337666666666666, "nominal_timestamp_seconds": 4.337666666666666, "decoded_png_sha256": "653cda1439d2d7846b9557eb3c580f498bb3d98c9083ed446b1284e5b0115a63", "rgb_array_sha256": "60e07074f9efc06140d020607017746a49f29e49314d09318bb655ca3666f57b" }, { "source_frame_index": 138, "source_pts": 138138, "timestamp_seconds": 4.6046, "nominal_timestamp_seconds": 4.6046, "decoded_png_sha256": "66a0d8c2e0abf866cdcef62b70d64c431b35bb976b807f92b21d5ffc9f6acc87", "rgb_array_sha256": "b47e9c09a42e5eb6ed7328e383f579f546ca315bc1add5c8738b6d672335fa87" } ], "command": [ "tools/ffmpeg-7.1-shared/bin/ffmpeg", "-v", "info", "-nostdin", "-threads", "1", "-i", "data/video/waterloo4k/ref/14.mp4", "-vf", "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "-frames:v", "3", "-fps_mode", "vfr", "-filter_threads", "1", "-threads", "1", "-y", "decode-cache/c24aed5fb0474b651304/%02d.png" ] }, { "lane": "A", "source_file": "data/video/waterloo4k/dist/VP9_14_1080p_3.mp4", "source_sha256": "5d7022a7ee56628aabf9d9b73b55e865ea9d9c2465df4c2957fbce22a669cfbe", "source_stream": { "codec_name": "vp9", "width": 1920, "height": 1080, "r_frame_rate": "30000/1001", "avg_frame_rate": "30000/1001", "time_base": "1/1000" }, "frame_indices_zero_based": [ 122, 130, 138 ], "display_canvas_pixels": [ 1920, 1080 ], "filter": "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "frames": [ { "source_frame_index": 122, "source_pts": 4071, "timestamp_seconds": 4.071, "nominal_timestamp_seconds": 4.070733333333333, "decoded_png_sha256": "a137bed054a5b00d0e48a86a632edfb23b0e53e68b7765e5a847af8d243aa3e6", "rgb_array_sha256": "afb4541028799e5fe30cc27144679afba677e6976e9673db2b48990c5d939af2" }, { "source_frame_index": 130, "source_pts": 4338, "timestamp_seconds": 4.338, "nominal_timestamp_seconds": 4.337666666666666, "decoded_png_sha256": "51e308af53a05a406036a21b8422bab5bedabf1ef82ea48cae2822351b854c32", "rgb_array_sha256": "9f75169448e2c6bc674c0d3e0e7482f10c80277b659a7303ff187f7e5d0a1cbd" }, { "source_frame_index": 138, "source_pts": 4605, "timestamp_seconds": 4.605, "nominal_timestamp_seconds": 4.6046, "decoded_png_sha256": "40e2afea4e00cb2309709c9cd167c20c6b41fd2917312200a5c2a289f746667c", "rgb_array_sha256": "aff371817f6f26813af836554cd194fd4557b93160b9e9afa60008e690925bb3" } ], "command": [ "tools/ffmpeg-7.1-shared/bin/ffmpeg", "-v", "info", "-nostdin", "-threads", "1", "-i", "data/video/waterloo4k/dist/VP9_14_1080p_3.mp4", "-vf", "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "-frames:v", "3", "-fps_mode", "vfr", "-filter_threads", "1", "-threads", "1", "-y", "decode-cache/c409bc5234b33c92e3df/%02d.png" ] }, { "lane": "B", "source_file": "data/video/waterloo4k/dist/H264_14_1080p_2.mp4", "source_sha256": "97081c2ba91616e1dd48815591c3b35d024b756db7761e6b6d3df4a745166a73", "source_stream": { "codec_name": "h264", "width": 1920, "height": 1080, "r_frame_rate": "30000/1001", "avg_frame_rate": "30000/1001", "time_base": "1/30000", "duration": "10.010000", "nb_frames": "300" }, "frame_indices_zero_based": [ 122, 130, 138 ], "display_canvas_pixels": [ 1920, 1080 ], "filter": "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "frames": [ { "source_frame_index": 122, "source_pts": 122122, "timestamp_seconds": 4.070733333333333, "nominal_timestamp_seconds": 4.070733333333333, "decoded_png_sha256": "4bc0d15194312f814c3164ecc5770724d5bae42e084894110db071347638fd41", "rgb_array_sha256": "999843a9f604ef7367618eb71fab5cccc1f983f074dfb90861a957ee07e62b95" }, { "source_frame_index": 130, "source_pts": 130130, "timestamp_seconds": 4.337666666666666, "nominal_timestamp_seconds": 4.337666666666666, "decoded_png_sha256": "45459f788df2024b8c97b112cecf2aa660c70febb1e8ca469fe332ad134e7d08", "rgb_array_sha256": "48b8223a2689ac35817f9a5f9725606d2ab56d290f357d2ede56b29e424e8390" }, { "source_frame_index": 138, "source_pts": 138138, "timestamp_seconds": 4.6046, "nominal_timestamp_seconds": 4.6046, "decoded_png_sha256": "3aa5f69ea62cc4fb402cc6882fbb55fce8548b654d3da4570caabcf1bbdbc410", "rgb_array_sha256": "a80e710a8f67172372211b4534f83833c6d178cc447b2b42ad10d46f400a0845" } ], "command": [ "tools/ffmpeg-7.1-shared/bin/ffmpeg", "-v", "info", "-nostdin", "-threads", "1", "-i", "data/video/waterloo4k/dist/H264_14_1080p_2.mp4", "-vf", "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "-frames:v", "3", "-fps_mode", "vfr", "-filter_threads", "1", "-threads", "1", "-y", "decode-cache/5fab32b49de65a2d82c4/%02d.png" ] } ], "scores": [ { "lane": "A", "reference": "data/video/waterloo4k/ref/14.mp4", "distortion": "data/video/waterloo4k/dist/VP9_14_1080p_3.mp4", "mos": 85.62, "lens_distance": 2.29745814647471, "vmaf": 57.883897, "source_file": "results/video2_pairs.csv", "source_line": 216, "source_columns": [ "mos", "videomae_b_k400_l2_lens_k16", "vmaf" ], "recorded_row": { "ref": "data/video/waterloo4k/ref/14.mp4", "dist": "data/video/waterloo4k/dist/VP9_14_1080p_3.mp4", "mos": "85.62", "dtype": "VP9", "level": "3", "split": "all", "vjepa2_l1_jld_k32": "11.46948081866105", "vjepa2_l1_jld_k64": "15.948037886039485", "vjepa2_l1_jld_k128": "20.535045162696488", "vjepa2_l2_jld_k32": "10.928199039565122", "vjepa2_l2_jld_k64": "15.662318349943304", "vjepa2_l2_jld_k128": "20.162364889733976", "vjepa2_l1_pca64": "20.69291639155394", "vjepa2_l2_pca64": "19.388702371034743", "vjepa2_l1_full1024": "28.59499135475675", "vjepa2_l2_full1024": "28.01127465635106", "psnr": "30.05467009544373", "ssim": "0.9131647571921347", "ms_ssim": "0.97771018743515", "image_jld_l1_k64": "0.1232373937964439", "vmaf": "57.883897", "config_hash": "9958b2bcc4", "git_commit": "a7a2254", "videomae_b_k400_l1_lens_k16": "2.782346112999047", "videomae_b_k400_l1_lens_k32": "4.002498770099477", "videomae_b_k400_l1_lens_k64": "5.6996431146961255", "videomae_b_k400_l1_lens_k128": "8.698138193442695", "videomae_b_k400_l2_lens_k16": "2.29745814647471", "videomae_b_k400_l2_lens_k32": "3.2925759439826607", "videomae_b_k400_l2_lens_k64": "4.441308818140991", "videomae_b_k400_l2_lens_k128": "6.2950414444435525", "videomae_b_k400_l1_pca_k64": "10.833219092329148", "videomae_b_k400_l2_pca_k64": "7.862883192467549", "videomae_b_k400_l1_full": "17.432375794196215", "videomae_b_k400_l2_full": "14.741755492850448", "video2_config_hash": "2d32247a64", "video2_git_commit": "ebe32e1", "videomae_b_ssl_l1_lens_k16": "3.094572227724843", "videomae_b_ssl_l1_lens_k32": "5.156867943928697", "videomae_b_ssl_l1_lens_k64": "8.072820991851577", "videomae_b_ssl_l1_lens_k128": "11.20192798300806", "videomae_b_ssl_l2_lens_k16": "2.729539912425874", "videomae_b_ssl_l2_lens_k32": "4.069892729861632", "videomae_b_ssl_l2_lens_k64": "6.118948418709213", "videomae_b_ssl_l2_lens_k128": "9.001205999316907", "videomae_b_ssl_l1_pca_k64": "13.63139347244031", "videomae_b_ssl_l2_pca_k64": "10.480761387081078", "videomae_b_ssl_l1_full": "19.97896203767311", "videomae_b_ssl_l2_full": "17.49534872095083", "vjepa2_l_predictor_l1_lens_k16": "7.55155859516342", "vjepa2_l_predictor_l1_lens_k32": "11.312038601902515", "vjepa2_l_predictor_l1_lens_k64": "15.728403937770912", "vjepa2_l_predictor_l1_lens_k128": "20.4299416287373", "vjepa2_l_predictor_l2_lens_k16": "7.380272516165803", "vjepa2_l_predictor_l2_lens_k32": "10.705796001760497", "vjepa2_l_predictor_l2_lens_k64": "15.263458754210316", "vjepa2_l_predictor_l2_lens_k128": "19.970865964342224", "vjepa2_l_predictor_l1_pca_k64": "20.69291639155394", "vjepa2_l_predictor_l2_pca_k64": "19.388702371034746", "vjepa2_l_predictor_l1_full": "28.59499135475675", "vjepa2_l_predictor_l2_full": "28.01127465635106", "source_line": 216 } }, { "lane": "B", "reference": "data/video/waterloo4k/ref/14.mp4", "distortion": "data/video/waterloo4k/dist/H264_14_1080p_2.mp4", "mos": 26.16, "lens_distance": 4.261681225241712, "vmaf": 59.29094, "source_file": "results/video2_pairs.csv", "source_line": 55, "source_columns": [ "mos", "videomae_b_k400_l2_lens_k16", "vmaf" ], "recorded_row": { "ref": "data/video/waterloo4k/ref/14.mp4", "dist": "data/video/waterloo4k/dist/H264_14_1080p_2.mp4", "mos": "26.16", "dtype": "H264", "level": "2", "split": "all", "vjepa2_l1_jld_k32": "20.114828441778656", "vjepa2_l1_jld_k64": "28.665025205709227", "vjepa2_l1_jld_k128": "38.47940503459488", "vjepa2_l2_jld_k32": "18.151091113540826", "vjepa2_l2_jld_k64": "26.238654745787795", "vjepa2_l2_jld_k128": "35.24446809256638", "vjepa2_l1_pca64": "41.74095905658933", "vjepa2_l2_pca64": "36.53222529640676", "vjepa2_l1_full1024": "55.78785986966959", "vjepa2_l2_full1024": "52.21403616375357", "psnr": "24.246647596359253", "ssim": "0.7214370369911194", "ms_ssim": "0.9058467820286752", "image_jld_l1_k64": "0.2426897846162319", "vmaf": "59.29094", "config_hash": "9958b2bcc4", "git_commit": "6653a35", "videomae_b_k400_l1_lens_k16": "5.788628174503387", "videomae_b_k400_l1_lens_k32": "8.299200475630077", "videomae_b_k400_l1_lens_k64": "11.63928513976517", "videomae_b_k400_l1_lens_k128": "16.842710048865687", "videomae_b_k400_l2_lens_k16": "4.261681225241712", "videomae_b_k400_l2_lens_k32": "6.106402871786487", "videomae_b_k400_l2_lens_k64": "8.2057522009007", "videomae_b_k400_l2_lens_k128": "11.290548154229972", "videomae_b_k400_l1_pca_k64": "19.67531506529695", "videomae_b_k400_l2_pca_k64": "14.351101854933772", "videomae_b_k400_l1_full": "30.087958518363028", "videomae_b_k400_l2_full": "24.45005206705178", "video2_config_hash": "2d32247a64", "video2_git_commit": "ebe32e1", "videomae_b_ssl_l1_lens_k16": "6.257927939961086", "videomae_b_ssl_l1_lens_k32": "10.28603532267422", "videomae_b_ssl_l1_lens_k64": "15.428059830584932", "videomae_b_ssl_l1_lens_k128": "20.295939106027014", "videomae_b_ssl_l2_lens_k16": "5.034791233276714", "videomae_b_ssl_l2_lens_k32": "7.3736850730105585", "videomae_b_ssl_l2_lens_k64": "10.749854774381284", "videomae_b_ssl_l2_lens_k128": "14.999897257135109", "videomae_b_ssl_l1_pca_k64": "23.755538937014", "videomae_b_ssl_l2_pca_k64": "17.743712265227135", "videomae_b_ssl_l1_full": "33.77458821122607", "videomae_b_ssl_l2_full": "28.16868325262158", "vjepa2_l_predictor_l1_lens_k16": "13.587093363329876", "vjepa2_l_predictor_l1_lens_k32": "20.17339841023244", "vjepa2_l_predictor_l1_lens_k64": "28.442512757884487", "vjepa2_l_predictor_l1_lens_k128": "38.238114249367165", "vjepa2_l_predictor_l2_lens_k16": "12.584408686753592", "vjepa2_l_predictor_l2_lens_k32": "18.025971220096142", "vjepa2_l_predictor_l2_lens_k64": "25.806962941087736", "vjepa2_l_predictor_l2_lens_k128": "34.871365921840265", "vjepa2_l_predictor_l1_pca_k64": "41.74095905658933", "vjepa2_l_predictor_l2_pca_k64": "36.53222529640676", "vjepa2_l_predictor_l1_full": "55.78785986966959", "vjepa2_l_predictor_l2_full": "52.21403616375357", "source_line": 55 } } ], "whole_video_preferences": { "MOS": "A", "Lens": "A", "VMAF": "B" }, "crop_xywh_on_1920x1080": [ 760, 760, 384, 216 ], "crop_selection": "Manually fixed display ROI after inspecting frame 130; identical spatial window for all three videos and all three times. No score is computed from this crop.", "artifact_annotation": { "text": "B: fine rail detail lost", "lanes": [ "B" ], "box_xywh": [ 110, 40, 120, 90 ], "frame_index": 130, "scope": "Qualitative observation in decoded pixels, not metric attribution, localisation truth or evidence of a temporal-quality mechanism." }, "original_v13_example": { "source": "14", "mos_gap": 59.46000000000001, "clips": [ { "ref": "data/video/waterloo4k/ref/14.mp4", "dist": "data/video/waterloo4k/dist/VP9_14_1080p_3.mp4", "mos": 85.62, "video_distance": 2.29745814647471, "vmaf": 57.883897, "source_file": "results/video2_pairs.csv", "source_line": 216, "source_columns": [ "mos", "videomae_b_k400_l2_lens_k16", "vmaf" ], "displayed": { "mos": "85.62", "video_distance": "2.30", "vmaf": "57.88" }, "frame": { "source_file": "data/video/waterloo4k/dist/VP9_14_1080p_3.mp4", "frame_index": 130, "index_origin": 0, "frame_selection": "Middle frame of the fourth evenly spaced evaluation clip.", "source_width": 1920, "source_height": 1080, "preview_file": "runtime/cache/check/v5_video_frames/d083b85a866f1ac1e171.png", "preview_sha256": "51e308af53a05a406036a21b8422bab5bedabf1ef82ea48cae2822351b854c32", "filter": "select=eq(n\\,130),scale=1920:1080:flags=lanczos", "command": [ "tools/ffmpeg-7.1-shared/bin/ffmpeg", "-v", "error", "-nostdin", "-threads", "1", "-i", "data/video/waterloo4k/dist/VP9_14_1080p_3.mp4", "-vf", "select=eq(n\\,130),scale=1920:1080:flags=lanczos", "-frames:v", "1", "-fps_mode", "vfr", "-filter_threads", "1", "-threads", "1", "-y", "runtime/cache/check/v5_video_frames/d083b85a866f1ac1e171.png" ], "crop": false } }, { "ref": "data/video/waterloo4k/ref/14.mp4", "dist": "data/video/waterloo4k/dist/H264_14_1080p_2.mp4", "mos": 26.16, "video_distance": 4.261681225241712, "vmaf": 59.29094, "source_file": "results/video2_pairs.csv", "source_line": 55, "source_columns": [ "mos", "videomae_b_k400_l2_lens_k16", "vmaf" ], "displayed": { "mos": "26.16", "video_distance": "4.26", "vmaf": "59.29" }, "frame": { "source_file": "data/video/waterloo4k/dist/H264_14_1080p_2.mp4", "frame_index": 130, "index_origin": 0, "frame_selection": "Middle frame of the fourth evenly spaced evaluation clip.", "source_width": 1920, "source_height": 1080, "preview_file": "runtime/cache/check/v5_video_frames/3e73c6d31178a24deb91.png", "preview_sha256": "45459f788df2024b8c97b112cecf2aa660c70febb1e8ca469fe332ad134e7d08", "filter": "select=eq(n\\,130),scale=1920:1080:flags=lanczos", "command": [ "tools/ffmpeg-7.1-shared/bin/ffmpeg", "-v", "error", "-nostdin", "-threads", "1", "-i", "data/video/waterloo4k/dist/H264_14_1080p_2.mp4", "-vf", "select=eq(n\\,130),scale=1920:1080:flags=lanczos", "-frames:v", "1", "-fps_mode", "vfr", "-filter_threads", "1", "-threads", "1", "-y", "runtime/cache/check/v5_video_frames/3e73c6d31178a24deb91.png" ], "crop": false } } ], "display_crop": { "top": 705, "left": 670, "width": 512, "height": 288, "canvas": [ 1920, 1080 ], "rule": "Same 512 by 288 crop for both videos, maximising RGB squared difference between their real sampled frames.", "source": "Decoded real frames; scripts/81_fig5_video_v5.py:plot_examples" } } }, { "source": "08", "label": "Football", "selection_rule": "Preserved v13 example: largest MOS reversal per distinct source where lens agrees and VMAF reverses.", "clips": [ { "lane": "Ref", "source_file": "data/video/waterloo4k/ref/08.mp4", "source_sha256": "86b235b7ab6ec062755496904ea01641ba218eb5a07e532f0a2b8406e615eda0", "source_stream": { "codec_name": "h264", "width": 3840, "height": 2160, "r_frame_rate": "30000/1001", "avg_frame_rate": "30000/1001", "time_base": "1/30000", "duration": "10.010000", "nb_frames": "300" }, "frame_indices_zero_based": [ 122, 130, 138 ], "display_canvas_pixels": [ 1920, 1080 ], "filter": "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "frames": [ { "source_frame_index": 122, "source_pts": 122122, "timestamp_seconds": 4.070733333333333, "nominal_timestamp_seconds": 4.070733333333333, "decoded_png_sha256": "70478637346de6d6f845d431d521131b12abeaf5104be3ab02af91fd9ce96275", "rgb_array_sha256": "066a01e3b3ec09656f4f8c0ca9e8fe0a734f871341fd28753188d456302908ad" }, { "source_frame_index": 130, "source_pts": 130130, "timestamp_seconds": 4.337666666666666, "nominal_timestamp_seconds": 4.337666666666666, "decoded_png_sha256": "17105500c9fe2c6bb05a9485e71ac36019cf84a0061805c0d940bbf0f4e0a9e2", "rgb_array_sha256": "38ac44b15d5002daf9870534143dd48aa9d9d17cc6d61d1263a941f069583c29" }, { "source_frame_index": 138, "source_pts": 138138, "timestamp_seconds": 4.6046, "nominal_timestamp_seconds": 4.6046, "decoded_png_sha256": "d3be02f86f0c0b0744f131e0f8ba262f92393dda23ff7984f99886502f411e9b", "rgb_array_sha256": "36c85d9542b34a81fa0cb74dbf39d427bce2a2cb3c3d75a72e9dc80c2997e0b6" } ], "command": [ "tools/ffmpeg-7.1-shared/bin/ffmpeg", "-v", "info", "-nostdin", "-threads", "1", "-i", "data/video/waterloo4k/ref/08.mp4", "-vf", "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "-frames:v", "3", "-fps_mode", "vfr", "-filter_threads", "1", "-threads", "1", "-y", "decode-cache/7a4f2f2b1fb215661e57/%02d.png" ] }, { "lane": "A", "source_file": "data/video/waterloo4k/dist/VP9_08_1080p_3.mp4", "source_sha256": "9373256c22ddf00ffd2fa85ce1490f7492614cfbcf23046ca880711db2267327", "source_stream": { "codec_name": "vp9", "width": 1920, "height": 1080, "r_frame_rate": "30000/1001", "avg_frame_rate": "30000/1001", "time_base": "1/1000" }, "frame_indices_zero_based": [ 122, 130, 138 ], "display_canvas_pixels": [ 1920, 1080 ], "filter": "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "frames": [ { "source_frame_index": 122, "source_pts": 4071, "timestamp_seconds": 4.071, "nominal_timestamp_seconds": 4.070733333333333, "decoded_png_sha256": "2d64ef6bec1c45e97792e2e0576def786762cba4a7b572b8a10aca652a394281", "rgb_array_sha256": "4468e42e84a1ac855c8f8108db6b65dd9421e02d5e958e4eba29743f798233ae" }, { "source_frame_index": 130, "source_pts": 4338, "timestamp_seconds": 4.338, "nominal_timestamp_seconds": 4.337666666666666, "decoded_png_sha256": "be9580230d91d4fca28d0f50436573eda14580af20c51a5e7f996deb8a7f1b69", "rgb_array_sha256": "fc8995c7e80b024d22a09b687e0f1184b696bed9358557682f1459d3850d94a7" }, { "source_frame_index": 138, "source_pts": 4605, "timestamp_seconds": 4.605, "nominal_timestamp_seconds": 4.6046, "decoded_png_sha256": "cfe8c108d3503bc4fe4d0673950ce6e259f6fa8eece1f49a3af9424219a60654", "rgb_array_sha256": "ce6e2514d163298a402df274b28a421c0554cb1bde9e8a1e61072749cc9e9740" } ], "command": [ "tools/ffmpeg-7.1-shared/bin/ffmpeg", "-v", "info", "-nostdin", "-threads", "1", "-i", "data/video/waterloo4k/dist/VP9_08_1080p_3.mp4", "-vf", "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "-frames:v", "3", "-fps_mode", "vfr", "-filter_threads", "1", "-threads", "1", "-y", "decode-cache/797f574d539cc5cfcdd1/%02d.png" ] }, { "lane": "B", "source_file": "data/video/waterloo4k/dist/H264_08_1080p_2.mp4", "source_sha256": "edca2d803f64f34b416106872b920bec2494f1b392fed7b64441a48fdcb8906c", "source_stream": { "codec_name": "h264", "width": 1920, "height": 1080, "r_frame_rate": "30000/1001", "avg_frame_rate": "30000/1001", "time_base": "1/30000", "duration": "10.010000", "nb_frames": "300" }, "frame_indices_zero_based": [ 122, 130, 138 ], "display_canvas_pixels": [ 1920, 1080 ], "filter": "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "frames": [ { "source_frame_index": 122, "source_pts": 122122, "timestamp_seconds": 4.070733333333333, "nominal_timestamp_seconds": 4.070733333333333, "decoded_png_sha256": "a11350333917b8f7756a9e901abbc7fcdfd464b532f3acde9ebf698651ed9d62", "rgb_array_sha256": "492834e278c00adb903a59f4772ccef22db7f2f7d00dea888ec4ac1eb4a7a749" }, { "source_frame_index": 130, "source_pts": 130130, "timestamp_seconds": 4.337666666666666, "nominal_timestamp_seconds": 4.337666666666666, "decoded_png_sha256": "91ba68401f06cf7273257559c6212674302accae5a18b6e68a667bad5a5f40af", "rgb_array_sha256": "548e910116923df0c8a2f31dc96d57b96fe80d9ba053530b4b445499ef0fd3e4" }, { "source_frame_index": 138, "source_pts": 138138, "timestamp_seconds": 4.6046, "nominal_timestamp_seconds": 4.6046, "decoded_png_sha256": "25a5098b02302072086307824aa794327e509b5a31e971cbc18b0656a5b1908e", "rgb_array_sha256": "1d0c6e6e34c5864ad429483df4923f810c1f56de430c5364ffaba2d7f2b1a0fa" } ], "command": [ "tools/ffmpeg-7.1-shared/bin/ffmpeg", "-v", "info", "-nostdin", "-threads", "1", "-i", "data/video/waterloo4k/dist/H264_08_1080p_2.mp4", "-vf", "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "-frames:v", "3", "-fps_mode", "vfr", "-filter_threads", "1", "-threads", "1", "-y", "decode-cache/ceb24e1ae4a8bdb186e1/%02d.png" ] } ], "scores": [ { "lane": "A", "reference": "data/video/waterloo4k/ref/08.mp4", "distortion": "data/video/waterloo4k/dist/VP9_08_1080p_3.mp4", "mos": 77.35, "lens_distance": 3.873523215406876, "vmaf": 43.260555, "source_file": "results/video2_pairs.csv", "source_line": 192, "source_columns": [ "mos", "videomae_b_k400_l2_lens_k16", "vmaf" ], "recorded_row": { "ref": "data/video/waterloo4k/ref/08.mp4", "dist": "data/video/waterloo4k/dist/VP9_08_1080p_3.mp4", "mos": "77.35", "dtype": "VP9", "level": "3", "split": "all", "vjepa2_l1_jld_k32": "9.384691798210524", "vjepa2_l1_jld_k64": "12.728702784367066", "vjepa2_l1_jld_k128": "16.728487405569457", "vjepa2_l2_jld_k32": "9.856119479885692", "vjepa2_l2_jld_k64": "14.170351428114", "vjepa2_l2_jld_k128": "18.786070092482177", "vjepa2_l1_pca64": "15.461331586552346", "vjepa2_l2_pca64": "22.90436482706088", "vjepa2_l1_full1024": "21.302119571372053", "vjepa2_l2_full1024": "29.56989697425189", "psnr": "29.72987174987793", "ssim": "0.8087153434753418", "ms_ssim": "0.9312992990016936", "image_jld_l1_k64": "0.1410237783566117", "vmaf": "43.260555", "config_hash": "9958b2bcc4", "git_commit": "a7a2254", "videomae_b_k400_l1_lens_k16": "4.1369556939092575", "videomae_b_k400_l1_lens_k32": "5.850064543017513", "videomae_b_k400_l1_lens_k64": "8.347842466555987", "videomae_b_k400_l1_lens_k128": "11.99260348614322", "videomae_b_k400_l2_lens_k16": "3.873523215406876", "videomae_b_k400_l2_lens_k32": "5.420221697788184", "videomae_b_k400_l2_lens_k64": "7.411368874612911", "videomae_b_k400_l2_lens_k128": "10.080460861736311", "videomae_b_k400_l1_pca_k64": "11.458080016139268", "videomae_b_k400_l2_pca_k64": "10.873374072017478", "videomae_b_k400_l1_full": "20.404414049076667", "videomae_b_k400_l2_full": "19.82565467782945", "video2_config_hash": "2d32247a64", "video2_git_commit": "ebe32e1", "videomae_b_ssl_l1_lens_k16": "4.092696411019651", "videomae_b_ssl_l1_lens_k32": "6.85874598927604", "videomae_b_ssl_l1_lens_k64": "10.25811065217283", "videomae_b_ssl_l1_lens_k128": "13.494227870803227", "videomae_b_ssl_l2_lens_k16": "3.900386926824032", "videomae_b_ssl_l2_lens_k32": "5.783387093289559", "videomae_b_ssl_l2_lens_k64": "8.335722558522587", "videomae_b_ssl_l2_lens_k128": "11.390808857505364", "videomae_b_ssl_l1_pca_k64": "13.033479678185234", "videomae_b_ssl_l2_pca_k64": "12.03036222278809", "videomae_b_ssl_l1_full": "21.45216757575141", "videomae_b_ssl_l2_full": "20.550343360723986", "vjepa2_l_predictor_l1_lens_k16": "6.431585502214531", "vjepa2_l_predictor_l1_lens_k32": "9.2840410396535", "vjepa2_l_predictor_l1_lens_k64": "12.59561620473188", "vjepa2_l_predictor_l1_lens_k128": "16.65099075623017", "vjepa2_l_predictor_l2_lens_k16": "6.97107950115276", "vjepa2_l_predictor_l2_lens_k32": "9.81419700478156", "vjepa2_l_predictor_l2_lens_k64": "13.942855763881981", "vjepa2_l_predictor_l2_lens_k128": "18.58576935110399", "vjepa2_l_predictor_l1_pca_k64": "15.461331586552346", "vjepa2_l_predictor_l2_pca_k64": "22.904364827060878", "vjepa2_l_predictor_l1_full": "21.302119571372053", "vjepa2_l_predictor_l2_full": "29.56989697425189", "source_line": 192 } }, { "lane": "B", "reference": "data/video/waterloo4k/ref/08.mp4", "distortion": "data/video/waterloo4k/dist/H264_08_1080p_2.mp4", "mos": 19.73, "lens_distance": 7.789758496195285, "vmaf": 45.053263, "source_file": "results/video2_pairs.csv", "source_line": 31, "source_columns": [ "mos", "videomae_b_k400_l2_lens_k16", "vmaf" ], "recorded_row": { "ref": "data/video/waterloo4k/ref/08.mp4", "dist": "data/video/waterloo4k/dist/H264_08_1080p_2.mp4", "mos": "19.73", "dtype": "H264", "level": "2", "split": "all", "vjepa2_l1_jld_k32": "13.602375004058578", "vjepa2_l1_jld_k64": "18.55034529783597", "vjepa2_l1_jld_k128": "24.545662192548463", "vjepa2_l2_jld_k32": "14.695192635679582", "vjepa2_l2_jld_k64": "20.58361103730249", "vjepa2_l2_jld_k128": "27.57584879615881", "vjepa2_l1_pca64": "24.371198230537185", "vjepa2_l2_pca64": "38.93949255582783", "vjepa2_l1_full1024": "31.77215403995875", "vjepa2_l2_full1024": "46.36823965805491", "psnr": "27.489983558654785", "ssim": "0.6432044021785259", "ms_ssim": "0.8183509260416031", "image_jld_l1_k64": "0.2509775068610906", "vmaf": "45.053263", "config_hash": "9958b2bcc4", "git_commit": "6653a35", "videomae_b_k400_l1_lens_k16": "8.065210640811811", "videomae_b_k400_l1_lens_k32": "10.606665870727214", "videomae_b_k400_l1_lens_k64": "13.948690737193235", "videomae_b_k400_l1_lens_k128": "18.25591349489143", "videomae_b_k400_l2_lens_k16": "7.789758496195285", "videomae_b_k400_l2_lens_k32": "10.200298250737166", "videomae_b_k400_l2_lens_k64": "13.253240440997429", "videomae_b_k400_l2_lens_k128": "16.929089812469602", "videomae_b_k400_l1_pca_k64": "17.56917230783156", "videomae_b_k400_l2_pca_k64": "19.57071237974073", "videomae_b_k400_l1_full": "27.901957387864723", "videomae_b_k400_l2_full": "29.31353575741204", "video2_config_hash": "2d32247a64", "video2_git_commit": "ebe32e1", "videomae_b_ssl_l1_lens_k16": "6.777336045846216", "videomae_b_ssl_l1_lens_k32": "10.694519994449593", "videomae_b_ssl_l1_lens_k64": "15.096916798474725", "videomae_b_ssl_l1_lens_k128": "19.316560781263828", "videomae_b_ssl_l2_lens_k16": "7.028943815318559", "videomae_b_ssl_l2_lens_k32": "9.887085002383683", "videomae_b_ssl_l2_lens_k64": "13.571530407688424", "videomae_b_ssl_l2_lens_k128": "17.53540327136517", "videomae_b_ssl_l1_pca_k64": "18.831536075053595", "videomae_b_ssl_l2_pca_k64": "20.09780297109952", "videomae_b_ssl_l1_full": "28.730102619714163", "videomae_b_ssl_l2_full": "29.394046922920225", "vjepa2_l_predictor_l1_lens_k16": "9.83014756391721", "vjepa2_l_predictor_l1_lens_k32": "13.573770191273512", "vjepa2_l_predictor_l1_lens_k64": "18.429416500878865", "vjepa2_l_predictor_l1_lens_k128": "24.46876574597394", "vjepa2_l_predictor_l2_lens_k16": "10.805270032425511", "vjepa2_l_predictor_l2_lens_k32": "14.728699449667568", "vjepa2_l_predictor_l2_lens_k64": "20.4052678992051", "vjepa2_l_predictor_l2_lens_k128": "27.26466395042473", "vjepa2_l_predictor_l1_pca_k64": "24.371198230537182", "vjepa2_l_predictor_l2_pca_k64": "38.93949255582783", "vjepa2_l_predictor_l1_full": "31.77215403995875", "vjepa2_l_predictor_l2_full": "46.36823965805491", "source_line": 31 } } ], "whole_video_preferences": { "MOS": "A", "Lens": "A", "VMAF": "B" }, "crop_xywh_on_1920x1080": [ 1024, 480, 384, 216 ], "crop_selection": "Manually fixed display ROI after inspecting frame 130; identical spatial window for all three videos and all three times. No score is computed from this crop.", "artifact_annotation": { "text": "B: block edges on the player", "lanes": [ "B" ], "box_xywh": [ 150, 60, 110, 110 ], "frame_index": 130, "scope": "Qualitative observation in decoded pixels, not metric attribution, localisation truth or evidence of a temporal-quality mechanism." }, "original_v13_example": { "source": "08", "mos_gap": 57.61999999999999, "clips": [ { "ref": "data/video/waterloo4k/ref/08.mp4", "dist": "data/video/waterloo4k/dist/VP9_08_1080p_3.mp4", "mos": 77.35, "video_distance": 3.873523215406876, "vmaf": 43.260555, "source_file": "results/video2_pairs.csv", "source_line": 192, "source_columns": [ "mos", "videomae_b_k400_l2_lens_k16", "vmaf" ], "displayed": { "mos": "77.35", "video_distance": "3.87", "vmaf": "43.26" }, "frame": { "source_file": "data/video/waterloo4k/dist/VP9_08_1080p_3.mp4", "frame_index": 130, "index_origin": 0, "frame_selection": "Middle frame of the fourth evenly spaced evaluation clip.", "source_width": 1920, "source_height": 1080, "preview_file": "runtime/cache/check/v5_video_frames/3b04dba380d127049305.png", "preview_sha256": "be9580230d91d4fca28d0f50436573eda14580af20c51a5e7f996deb8a7f1b69", "filter": "select=eq(n\\,130),scale=1920:1080:flags=lanczos", "command": [ "tools/ffmpeg-7.1-shared/bin/ffmpeg", "-v", "error", "-nostdin", "-threads", "1", "-i", "data/video/waterloo4k/dist/VP9_08_1080p_3.mp4", "-vf", "select=eq(n\\,130),scale=1920:1080:flags=lanczos", "-frames:v", "1", "-fps_mode", "vfr", "-filter_threads", "1", "-threads", "1", "-y", "runtime/cache/check/v5_video_frames/3b04dba380d127049305.png" ], "crop": false } }, { "ref": "data/video/waterloo4k/ref/08.mp4", "dist": "data/video/waterloo4k/dist/H264_08_1080p_2.mp4", "mos": 19.73, "video_distance": 7.789758496195285, "vmaf": 45.053263, "source_file": "results/video2_pairs.csv", "source_line": 31, "source_columns": [ "mos", "videomae_b_k400_l2_lens_k16", "vmaf" ], "displayed": { "mos": "19.73", "video_distance": "7.79", "vmaf": "45.05" }, "frame": { "source_file": "data/video/waterloo4k/dist/H264_08_1080p_2.mp4", "frame_index": 130, "index_origin": 0, "frame_selection": "Middle frame of the fourth evenly spaced evaluation clip.", "source_width": 1920, "source_height": 1080, "preview_file": "runtime/cache/check/v5_video_frames/bcce3f579d13ca5f684c.png", "preview_sha256": "91ba68401f06cf7273257559c6212674302accae5a18b6e68a667bad5a5f40af", "filter": "select=eq(n\\,130),scale=1920:1080:flags=lanczos", "command": [ "tools/ffmpeg-7.1-shared/bin/ffmpeg", "-v", "error", "-nostdin", "-threads", "1", "-i", "data/video/waterloo4k/dist/H264_08_1080p_2.mp4", "-vf", "select=eq(n\\,130),scale=1920:1080:flags=lanczos", "-frames:v", "1", "-fps_mode", "vfr", "-filter_threads", "1", "-threads", "1", "-y", "runtime/cache/check/v5_video_frames/bcce3f579d13ca5f684c.png" ], "crop": false } } ], "display_crop": { "top": 484, "left": 964, "width": 512, "height": 288, "canvas": [ 1920, 1080 ], "rule": "Same 512 by 288 crop for both videos, maximising RGB squared difference between their real sampled frames.", "source": "Decoded real frames; scripts/81_fig5_video_v5.py:plot_examples" } } }, { "source": "16", "label": "Dance", "selection_rule": "Additional agreement: largest positive MOS gap among remaining sources with the stated metric ordering; ties break by A then B source path.", "clips": [ { "lane": "Ref", "source_file": "data/video/waterloo4k/ref/16.mp4", "source_sha256": "2db332e29c1f4cf86f43a3fb44ef15d9949bdd1abe10dfcf7d520ce384e61b5c", "source_stream": { "codec_name": "h264", "width": 3840, "height": 2160, "r_frame_rate": "30000/1001", "avg_frame_rate": "30000/1001", "time_base": "1/30000", "duration": "10.010000", "nb_frames": "300" }, "frame_indices_zero_based": [ 122, 130, 138 ], "display_canvas_pixels": [ 1920, 1080 ], "filter": "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "frames": [ { "source_frame_index": 122, "source_pts": 122122, "timestamp_seconds": 4.070733333333333, "nominal_timestamp_seconds": 4.070733333333333, "decoded_png_sha256": "dd54a3045a2ab844191bd2648b191befa42091fe79c245528049d31cbed2643e", "rgb_array_sha256": "a1e7070089bf8de9ff33f88146ce15dcdc9e15dd344ef621081bb7a14699fefa" }, { "source_frame_index": 130, "source_pts": 130130, "timestamp_seconds": 4.337666666666666, "nominal_timestamp_seconds": 4.337666666666666, "decoded_png_sha256": "c94b356581d572ef90ae834f2a80740bdf1cb29a5091f1c3825d688ae522ec88", "rgb_array_sha256": "7fb2930bfe330e7c6c1bf281c39ee84737ef1ef5c6901710ead30911db0ad613" }, { "source_frame_index": 138, "source_pts": 138138, "timestamp_seconds": 4.6046, "nominal_timestamp_seconds": 4.6046, "decoded_png_sha256": "7aa31792c508ae4fea9402b687c5c64e23bbdcc9efb7a838e5c1de6159cfab23", "rgb_array_sha256": "c2670b37d2dd11cdd12f93b17e3795888047934f4300b04494d7d7b501a5c2d1" } ], "command": [ "tools/ffmpeg-7.1-shared/bin/ffmpeg", "-v", "info", "-nostdin", "-threads", "1", "-i", "data/video/waterloo4k/ref/16.mp4", "-vf", "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "-frames:v", "3", "-fps_mode", "vfr", "-filter_threads", "1", "-threads", "1", "-y", "decode-cache/164bb3b58c0aab1b8888/%02d.png" ] }, { "lane": "A", "source_file": "data/video/waterloo4k/dist/VP9_16_1080p_3.mp4", "source_sha256": "9b4147574f6ab44e32e78a4db94ae9f8f90f7ee9390acef50b5ef2ad597f1dce", "source_stream": { "codec_name": "vp9", "width": 1920, "height": 1080, "r_frame_rate": "30000/1001", "avg_frame_rate": "30000/1001", "time_base": "1/1000" }, "frame_indices_zero_based": [ 122, 130, 138 ], "display_canvas_pixels": [ 1920, 1080 ], "filter": "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "frames": [ { "source_frame_index": 122, "source_pts": 4071, "timestamp_seconds": 4.071, "nominal_timestamp_seconds": 4.070733333333333, "decoded_png_sha256": "c002214cc05a6a38def774481a2b3896865abcba5688e0ed9bbea18df6b8c453", "rgb_array_sha256": "96be82677cafbc1e31468d74aad297da0face1d71a2b8f7fd7b24b9fbca403bc" }, { "source_frame_index": 130, "source_pts": 4338, "timestamp_seconds": 4.338, "nominal_timestamp_seconds": 4.337666666666666, "decoded_png_sha256": "06f94c25943a5c8c83c29e89de67a0246fc106eb60a3ec08b3ceafa908ca7ef5", "rgb_array_sha256": "3dddc5cae2ba0335396cab6d655b585b77103dfad55c8b43d00c96d9e60e94a3" }, { "source_frame_index": 138, "source_pts": 4605, "timestamp_seconds": 4.605, "nominal_timestamp_seconds": 4.6046, "decoded_png_sha256": "9ab2147a02e40829affa8adc972b028a97c608311dbef7b06fb3897169e8d6d6", "rgb_array_sha256": "41cb896a0ce5355de5cf3b5b1fd16cd0241bd6d49b03a2270072dc695f0a82c9" } ], "command": [ "tools/ffmpeg-7.1-shared/bin/ffmpeg", "-v", "info", "-nostdin", "-threads", "1", "-i", "data/video/waterloo4k/dist/VP9_16_1080p_3.mp4", "-vf", "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "-frames:v", "3", "-fps_mode", "vfr", "-filter_threads", "1", "-threads", "1", "-y", "decode-cache/7de557f4d4871541844d/%02d.png" ] }, { "lane": "B", "source_file": "data/video/waterloo4k/dist/HEVC_16_1080p_2.mp4", "source_sha256": "35d142fd1379e263022b954ca7b57c563239fd7841cf9dfccd42f09ffb86ebe9", "source_stream": { "codec_name": "hevc", "width": 1920, "height": 1080, "r_frame_rate": "30000/1001", "avg_frame_rate": "30000/1001", "time_base": "1/30000", "duration": "10.010000", "nb_frames": "300" }, "frame_indices_zero_based": [ 122, 130, 138 ], "display_canvas_pixels": [ 1920, 1080 ], "filter": "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "frames": [ { "source_frame_index": 122, "source_pts": 122122, "timestamp_seconds": 4.070733333333333, "nominal_timestamp_seconds": 4.070733333333333, "decoded_png_sha256": "93d28b0c86890ebe3c99bfcd056efa1e7fb2e20c3e977b51cf513f0805358069", "rgb_array_sha256": "9640ee181f75425512ce944f128b3f6adc68cd37fee425035108097730b5bb36" }, { "source_frame_index": 130, "source_pts": 130130, "timestamp_seconds": 4.337666666666666, "nominal_timestamp_seconds": 4.337666666666666, "decoded_png_sha256": "cffaa8940889757b55c8e5793684e2031feec3a0ba0dc14555c8779dc8eed2b7", "rgb_array_sha256": "a8ba8d08ddad40da7b2cb4959e320e103849dbdd6cbd64dfdb1406bb9065ecf6" }, { "source_frame_index": 138, "source_pts": 138138, "timestamp_seconds": 4.6046, "nominal_timestamp_seconds": 4.6046, "decoded_png_sha256": "f5afb7a7c1e02f70b89c584874ed1cd3e2c4e6ed4002775a7c8b2db26f62b5df", "rgb_array_sha256": "e5807c681ab55ad98a61781ed60095dccb8d7e6119ca3fbe48610a5325030322" } ], "command": [ "tools/ffmpeg-7.1-shared/bin/ffmpeg", "-v", "info", "-nostdin", "-threads", "1", "-i", "data/video/waterloo4k/dist/HEVC_16_1080p_2.mp4", "-vf", "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "-frames:v", "3", "-fps_mode", "vfr", "-filter_threads", "1", "-threads", "1", "-y", "decode-cache/9d6e760f631363808a61/%02d.png" ] } ], "scores": [ { "lane": "A", "reference": "data/video/waterloo4k/ref/16.mp4", "distortion": "data/video/waterloo4k/dist/VP9_16_1080p_3.mp4", "mos": 78.02, "lens_distance": 3.380661930553204, "vmaf": 60.630977, "source_file": "results/video2_pairs.csv", "source_line": 224, "source_columns": [ "mos", "videomae_b_k400_l2_lens_k16", "vmaf" ], "recorded_row": { "ref": "data/video/waterloo4k/ref/16.mp4", "dist": "data/video/waterloo4k/dist/VP9_16_1080p_3.mp4", "mos": "78.02", "dtype": "VP9", "level": "3", "split": "all", "vjepa2_l1_jld_k32": "11.707077031369026", "vjepa2_l1_jld_k64": "16.262224089197346", "vjepa2_l1_jld_k128": "21.48271229361493", "vjepa2_l2_jld_k32": "11.908146424336062", "vjepa2_l2_jld_k64": "16.80696820433589", "vjepa2_l2_jld_k128": "21.816087925022096", "vjepa2_l1_pca64": "19.327992468882982", "vjepa2_l2_pca64": "19.21691081853352", "vjepa2_l1_full1024": "26.928816030648285", "vjepa2_l2_full1024": "27.803197372220563", "psnr": "28.849427700042725", "ssim": "0.8891748115420341", "ms_ssim": "0.9584046825766565", "image_jld_l1_k64": "0.1616401690989732", "vmaf": "60.630977", "config_hash": "9958b2bcc4", "git_commit": "a7a2254", "videomae_b_k400_l1_lens_k16": "3.744611444854117", "videomae_b_k400_l1_lens_k32": "5.496585739878657", "videomae_b_k400_l1_lens_k64": "7.500131256226535", "videomae_b_k400_l1_lens_k128": "10.289524957203303", "videomae_b_k400_l2_lens_k16": "3.380661930553204", "videomae_b_k400_l2_lens_k32": "4.771380966383867", "videomae_b_k400_l2_lens_k64": "6.556959410943945", "videomae_b_k400_l2_lens_k128": "8.805986093718504", "videomae_b_k400_l1_pca_k64": "9.8499488809876", "videomae_b_k400_l2_pca_k64": "8.87316295254802", "videomae_b_k400_l1_full": "16.987206694124872", "videomae_b_k400_l2_full": "16.425907993621447", "video2_config_hash": "2d32247a64", "video2_git_commit": "ebe32e1", "videomae_b_ssl_l1_lens_k16": "3.893357998386621", "videomae_b_ssl_l1_lens_k32": "6.206341618133451", "videomae_b_ssl_l1_lens_k64": "9.03727637260941", "videomae_b_ssl_l1_lens_k128": "12.02363468278654", "videomae_b_ssl_l2_lens_k16": "3.757332610578876", "videomae_b_ssl_l2_lens_k32": "5.5664310709082025", "videomae_b_ssl_l2_lens_k64": "7.933207334858889", "videomae_b_ssl_l2_lens_k128": "10.850356010889657", "videomae_b_ssl_l1_pca_k64": "11.628354644763151", "videomae_b_ssl_l2_pca_k64": "10.370233199055034", "videomae_b_ssl_l1_full": "18.989584125707808", "videomae_b_ssl_l2_full": "18.35247231441229", "vjepa2_l_predictor_l1_lens_k16": "8.13045142765643", "vjepa2_l_predictor_l1_lens_k32": "11.529298984295398", "vjepa2_l_predictor_l1_lens_k64": "16.030565927644954", "vjepa2_l_predictor_l1_lens_k128": "21.395150307138277", "vjepa2_l_predictor_l2_lens_k16": "8.055000010513504", "vjepa2_l_predictor_l2_lens_k32": "11.626777554043025", "vjepa2_l_predictor_l2_lens_k64": "16.41500353578128", "vjepa2_l_predictor_l2_lens_k128": "21.602789084185872", "vjepa2_l_predictor_l1_pca_k64": "19.327992468882982", "vjepa2_l_predictor_l2_pca_k64": "19.21691081853352", "vjepa2_l_predictor_l1_full": "26.928816030648285", "vjepa2_l_predictor_l2_full": "27.803197372220563", "source_line": 224 } }, { "lane": "B", "reference": "data/video/waterloo4k/ref/16.mp4", "distortion": "data/video/waterloo4k/dist/HEVC_16_1080p_2.mp4", "mos": 23.27, "lens_distance": 5.973954513091839, "vmaf": 60.682214, "source_file": "results/video2_pairs.csv", "source_line": 143, "source_columns": [ "mos", "videomae_b_k400_l2_lens_k16", "vmaf" ], "recorded_row": { "ref": "data/video/waterloo4k/ref/16.mp4", "dist": "data/video/waterloo4k/dist/HEVC_16_1080p_2.mp4", "mos": "23.27", "dtype": "HEVC", "level": "2", "split": "all", "vjepa2_l1_jld_k32": "21.70063898226892", "vjepa2_l1_jld_k64": "30.38782459683528", "vjepa2_l1_jld_k128": "40.48363003915856", "vjepa2_l2_jld_k32": "20.892428860216302", "vjepa2_l2_jld_k64": "29.135311544313023", "vjepa2_l2_jld_k128": "38.3529307373319", "vjepa2_l1_pca64": "39.41566968198679", "vjepa2_l2_pca64": "36.85477547208308", "vjepa2_l1_full1024": "51.42272099112278", "vjepa2_l2_full1024": "50.4559830806783", "psnr": "23.3548424243927", "ssim": "0.7441339194774628", "ms_ssim": "0.8827340453863144", "image_jld_l1_k64": "0.2972599677741527", "vmaf": "60.682214", "config_hash": "9958b2bcc4", "git_commit": "6653a35", "videomae_b_k400_l1_lens_k16": "7.609997295023253", "videomae_b_k400_l1_lens_k32": "10.617072929342395", "videomae_b_k400_l1_lens_k64": "13.842615154242774", "videomae_b_k400_l1_lens_k128": "18.074560870776", "videomae_b_k400_l2_lens_k16": "5.973954513091839", "videomae_b_k400_l2_lens_k32": "8.211884477561837", "videomae_b_k400_l2_lens_k64": "10.912020842529484", "videomae_b_k400_l2_lens_k128": "14.172045655153592", "videomae_b_k400_l1_pca_k64": "17.523544113916508", "videomae_b_k400_l2_pca_k64": "15.002485768661526", "videomae_b_k400_l1_full": "28.02774675777329", "videomae_b_k400_l2_full": "25.172391008881217", "video2_config_hash": "2d32247a64", "video2_git_commit": "ebe32e1", "videomae_b_ssl_l1_lens_k16": "7.21474659538843", "videomae_b_ssl_l1_lens_k32": "11.068585074610668", "videomae_b_ssl_l1_lens_k64": "15.51908874264922", "videomae_b_ssl_l1_lens_k128": "20.15864659928448", "videomae_b_ssl_l2_lens_k16": "6.337619318393771", "videomae_b_ssl_l2_lens_k32": "9.130919117451612", "videomae_b_ssl_l2_lens_k64": "12.730828775545442", "videomae_b_ssl_l2_lens_k128": "16.98966834325551", "videomae_b_ssl_l1_pca_k64": "19.801358721574108", "videomae_b_ssl_l2_pca_k64": "17.02468734431716", "videomae_b_ssl_l1_full": "30.739729740143623", "videomae_b_ssl_l2_full": "28.03222464033372", "vjepa2_l_predictor_l1_lens_k16": "15.526734399894845", "vjepa2_l_predictor_l1_lens_k32": "21.567310183719968", "vjepa2_l_predictor_l1_lens_k64": "30.03118420136197", "vjepa2_l_predictor_l1_lens_k128": "40.313362999483495", "vjepa2_l_predictor_l2_lens_k16": "14.46505676986256", "vjepa2_l_predictor_l2_lens_k32": "20.50156238463746", "vjepa2_l_predictor_l2_lens_k64": "28.628924504744145", "vjepa2_l_predictor_l2_lens_k128": "37.94686386961461", "vjepa2_l_predictor_l1_pca_k64": "39.415669681986785", "vjepa2_l_predictor_l2_pca_k64": "36.85477547208308", "vjepa2_l_predictor_l1_full": "51.42272099112278", "vjepa2_l_predictor_l2_full": "50.4559830806783", "source_line": 143 } } ], "whole_video_preferences": { "MOS": "A", "Lens": "A", "VMAF": "B" }, "crop_xywh_on_1920x1080": [ 960, 340, 384, 216 ], "crop_selection": "Manually fixed display ROI after inspecting frame 130; identical spatial window for all three videos and all three times. No score is computed from this crop.", "artifact_annotation": { "text": "B: facial detail lost", "lanes": [ "B" ], "box_xywh": [ 16, 8, 108, 112 ], "frame_index": 130, "scope": "Qualitative observation in decoded pixels, not metric attribution, localisation truth or evidence of a temporal-quality mechanism." }, "original_v13_example": null }, { "source": "11", "label": "Game scene", "selection_rule": "Additional failure: largest positive MOS gap among remaining sources with the stated metric ordering; ties break by A then B source path.", "clips": [ { "lane": "Ref", "source_file": "data/video/waterloo4k/ref/11.mp4", "source_sha256": "dcbd82d85e0a33b09c506a9b76a968013f104c927533ac001e6819ae3809c912", "source_stream": { "codec_name": "h264", "width": 3840, "height": 2160, "r_frame_rate": "30/1", "avg_frame_rate": "30/1", "time_base": "1/30000", "duration": "10.000000", "nb_frames": "300" }, "frame_indices_zero_based": [ 122, 130, 138 ], "display_canvas_pixels": [ 1920, 1080 ], "filter": "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "frames": [ { "source_frame_index": 122, "source_pts": 122000, "timestamp_seconds": 4.066666666666666, "nominal_timestamp_seconds": 4.066666666666666, "decoded_png_sha256": "774b0bfaa69bc8193daea01fd23c97c811720356cec997fbbbaf85fd8b6ddb65", "rgb_array_sha256": "d6c42da73cee35a7c5bf3d2a8bf31a49f9f0406c378bb07d558c28ea87ebc88b" }, { "source_frame_index": 130, "source_pts": 130000, "timestamp_seconds": 4.333333333333333, "nominal_timestamp_seconds": 4.333333333333333, "decoded_png_sha256": "05cba06fbff44ed071361b20e1fd3c5ef9937bc3b374a9c52cff5d81bb723e85", "rgb_array_sha256": "baf207e7e52e8f1822a93dc1d039db231beff1040f6ddd4e09fd4eaf30017873" }, { "source_frame_index": 138, "source_pts": 138000, "timestamp_seconds": 4.6, "nominal_timestamp_seconds": 4.6, "decoded_png_sha256": "0b15e2b176208e394f4084c6f3ff938f3dbd7aa95a0cea4b00c651249f8f5d6f", "rgb_array_sha256": "28f42f6dc691363234d75ec76fe1a2e284290e81fcc51db0afdb291ca3894f1f" } ], "command": [ "tools/ffmpeg-7.1-shared/bin/ffmpeg", "-v", "info", "-nostdin", "-threads", "1", "-i", "data/video/waterloo4k/ref/11.mp4", "-vf", "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "-frames:v", "3", "-fps_mode", "vfr", "-filter_threads", "1", "-threads", "1", "-y", "decode-cache/9daf0fe845cf0233e44d/%02d.png" ] }, { "lane": "A", "source_file": "data/video/waterloo4k/dist/HEVC_11_1080p_3.mp4", "source_sha256": "996ba236cd946b047619805e6f7de41ddbbec17eb1bd413f2874f5c90716e8d3", "source_stream": { "codec_name": "hevc", "width": 1920, "height": 1080, "r_frame_rate": "30/1", "avg_frame_rate": "30/1", "time_base": "1/15360", "duration": "10.000000", "nb_frames": "300" }, "frame_indices_zero_based": [ 122, 130, 138 ], "display_canvas_pixels": [ 1920, 1080 ], "filter": "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "frames": [ { "source_frame_index": 122, "source_pts": 62464, "timestamp_seconds": 4.066666666666666, "nominal_timestamp_seconds": 4.066666666666666, "decoded_png_sha256": "b6e3a3a19a7f1ee6e709efa44e42bf9591bb3987a88ce1b694428c59501d3664", "rgb_array_sha256": "cae7d8513364baeec51c873e1835248420286dd83a7af244c7044d8a73021525" }, { "source_frame_index": 130, "source_pts": 66560, "timestamp_seconds": 4.333333333333333, "nominal_timestamp_seconds": 4.333333333333333, "decoded_png_sha256": "f84343f5c9c28039bce57ba365c69bd00479898822661b81974fa8e4e780118b", "rgb_array_sha256": "c03a21ebc89a1f8f27ae8d8f18c67f35a5956187885affffcc7a02331d4eaec1" }, { "source_frame_index": 138, "source_pts": 70656, "timestamp_seconds": 4.6, "nominal_timestamp_seconds": 4.6, "decoded_png_sha256": "a1b0824dded631b10f0f2b75e79f43f2f25b477a8b5e7606c48937436793ec59", "rgb_array_sha256": "6c69b376c52decbe910a10c1b204c9d7c6d15f6c04cf95fe892e73374d8ed4a4" } ], "command": [ "tools/ffmpeg-7.1-shared/bin/ffmpeg", "-v", "info", "-nostdin", "-threads", "1", "-i", "data/video/waterloo4k/dist/HEVC_11_1080p_3.mp4", "-vf", "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "-frames:v", "3", "-fps_mode", "vfr", "-filter_threads", "1", "-threads", "1", "-y", "decode-cache/9e6ec2c63a3e01d76869/%02d.png" ] }, { "lane": "B", "source_file": "data/video/waterloo4k/dist/H264_11_1080p_3.mp4", "source_sha256": "d7b1f1b38ccc5848a6bc7e6299b6ab048adb317fedd3eecf3993495ffdb5a00b", "source_stream": { "codec_name": "h264", "width": 1920, "height": 1080, "r_frame_rate": "30/1", "avg_frame_rate": "30/1", "time_base": "1/15360", "duration": "10.000000", "nb_frames": "300" }, "frame_indices_zero_based": [ 122, 130, 138 ], "display_canvas_pixels": [ 1920, 1080 ], "filter": "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "frames": [ { "source_frame_index": 122, "source_pts": 62464, "timestamp_seconds": 4.066666666666666, "nominal_timestamp_seconds": 4.066666666666666, "decoded_png_sha256": "05819531ee4273226e27f3e08dadfd9d02bfaab17aeea12d415a419eaac3f3a5", "rgb_array_sha256": "f4db5bcd5b74e2af6afc96b257adf9b1e660d90a45926342a1af213dcec6ff0a" }, { "source_frame_index": 130, "source_pts": 66560, "timestamp_seconds": 4.333333333333333, "nominal_timestamp_seconds": 4.333333333333333, "decoded_png_sha256": "d6592a9e78bd89da4f956f8c466773e3b961ac324f350f0707ddce578834f0c4", "rgb_array_sha256": "62b1c916a0e64329a08431c551a60078436192c61074c4b65b0643881c077049" }, { "source_frame_index": 138, "source_pts": 70656, "timestamp_seconds": 4.6, "nominal_timestamp_seconds": 4.6, "decoded_png_sha256": "475b7caefd04a4de3ee7b5c558270c7d51e31176e8b1b6cdd39b68476619b459", "rgb_array_sha256": "6fb2c1ba29a8bee1563eaaf282cc574aa408afafa4bc158bf1cfa88e3c84d656" } ], "command": [ "tools/ffmpeg-7.1-shared/bin/ffmpeg", "-v", "info", "-nostdin", "-threads", "1", "-i", "data/video/waterloo4k/dist/H264_11_1080p_3.mp4", "-vf", "select=eq(n\\,122)+eq(n\\,130)+eq(n\\,138),showinfo,scale=1920:1080:flags=lanczos", "-frames:v", "3", "-fps_mode", "vfr", "-filter_threads", "1", "-threads", "1", "-y", "decode-cache/784f697bbc0899a131ff/%02d.png" ] } ], "scores": [ { "lane": "A", "reference": "data/video/waterloo4k/ref/11.mp4", "distortion": "data/video/waterloo4k/dist/HEVC_11_1080p_3.mp4", "mos": 58.42, "lens_distance": 4.815775140777205, "vmaf": 59.615028, "source_file": "results/video2_pairs.csv", "source_line": 124, "source_columns": [ "mos", "videomae_b_k400_l2_lens_k16", "vmaf" ], "recorded_row": { "ref": "data/video/waterloo4k/ref/11.mp4", "dist": "data/video/waterloo4k/dist/HEVC_11_1080p_3.mp4", "mos": "58.42", "dtype": "HEVC", "level": "3", "split": "all", "vjepa2_l1_jld_k32": "10.656841496582771", "vjepa2_l1_jld_k64": "14.851772734284284", "vjepa2_l1_jld_k128": "20.04203927162312", "vjepa2_l2_jld_k32": "11.719017412313518", "vjepa2_l2_jld_k64": "16.41254535328859", "vjepa2_l2_jld_k128": "21.73984860701782", "vjepa2_l1_pca64": "18.99760050429144", "vjepa2_l2_pca64": "20.60055799932184", "vjepa2_l1_full1024": "25.74222795544052", "vjepa2_l2_full1024": "28.501370179551305", "psnr": "29.953941106796265", "ssim": "0.8379626721143723", "ms_ssim": "0.9275333881378174", "image_jld_l1_k64": "0.1956483889371156", "vmaf": "59.615028", "config_hash": "9958b2bcc4", "git_commit": "6653a35", "videomae_b_k400_l1_lens_k16": "5.239137242503925", "videomae_b_k400_l1_lens_k32": "7.657339913777499", "videomae_b_k400_l1_lens_k64": "10.33354591848266", "videomae_b_k400_l1_lens_k128": "13.56610481417203", "videomae_b_k400_l2_lens_k16": "4.815775140777205", "videomae_b_k400_l2_lens_k32": "6.739325984736615", "videomae_b_k400_l2_lens_k64": "9.134524091194878", "videomae_b_k400_l2_lens_k128": "11.990448925722866", "videomae_b_k400_l1_pca_k64": "11.775810847944037", "videomae_b_k400_l2_pca_k64": "11.976447117915953", "videomae_b_k400_l1_full": "19.855227057475474", "videomae_b_k400_l2_full": "20.729086206430164", "video2_config_hash": "2d32247a64", "video2_git_commit": "ebe32e1", "videomae_b_ssl_l1_lens_k16": "4.885207371678354", "videomae_b_ssl_l1_lens_k32": "7.869131520470309", "videomae_b_ssl_l1_lens_k64": "11.24769816691876", "videomae_b_ssl_l1_lens_k128": "14.3220998561187", "videomae_b_ssl_l2_lens_k16": "4.750226994160171", "videomae_b_ssl_l2_lens_k32": "6.95836402596959", "videomae_b_ssl_l2_lens_k64": "9.82893026233927", "videomae_b_ssl_l2_lens_k128": "13.079112062383787", "videomae_b_ssl_l1_pca_k64": "12.813720533340629", "videomae_b_ssl_l2_pca_k64": "12.421474196007791", "videomae_b_ssl_l1_full": "20.6670230814236", "videomae_b_ssl_l2_full": "21.06991175071365", "vjepa2_l_predictor_l1_lens_k16": "7.742317483467145", "vjepa2_l_predictor_l1_lens_k32": "10.618706713089216", "vjepa2_l_predictor_l1_lens_k64": "14.741374904634219", "vjepa2_l_predictor_l1_lens_k128": "19.958491978449498", "vjepa2_l_predictor_l2_lens_k16": "8.17467361871916", "vjepa2_l_predictor_l2_lens_k32": "11.541880228061352", "vjepa2_l_predictor_l2_lens_k64": "16.163892875065304", "vjepa2_l_predictor_l2_lens_k128": "21.505139025035017", "vjepa2_l_predictor_l1_pca_k64": "18.997600504291437", "vjepa2_l_predictor_l2_pca_k64": "20.600557999321836", "vjepa2_l_predictor_l1_full": "25.742227955440516", "vjepa2_l_predictor_l2_full": "28.501370179551305", "source_line": 124 } }, { "lane": "B", "reference": "data/video/waterloo4k/ref/11.mp4", "distortion": "data/video/waterloo4k/dist/H264_11_1080p_3.mp4", "mos": 43.47, "lens_distance": 4.790580446716685, "vmaf": 54.63287, "source_file": "results/video2_pairs.csv", "source_line": 44, "source_columns": [ "mos", "videomae_b_k400_l2_lens_k16", "vmaf" ], "recorded_row": { "ref": "data/video/waterloo4k/ref/11.mp4", "dist": "data/video/waterloo4k/dist/H264_11_1080p_3.mp4", "mos": "43.47", "dtype": "H264", "level": "3", "split": "all", "vjepa2_l1_jld_k32": "11.385734125558676", "vjepa2_l1_jld_k64": "15.66741402175903", "vjepa2_l1_jld_k128": "21.018735566078966", "vjepa2_l2_jld_k32": "12.047339246144029", "vjepa2_l2_jld_k64": "16.885883410351195", "vjepa2_l2_jld_k128": "22.420015456018216", "vjepa2_l1_pca64": "19.457112662714263", "vjepa2_l2_pca64": "20.69424669596544", "vjepa2_l1_full1024": "26.777283133251046", "vjepa2_l2_full1024": "29.16930781440626", "psnr": "29.30282974243164", "ssim": "0.8296336680650711", "ms_ssim": "0.9218063578009604", "image_jld_l1_k64": "0.2020241990685463", "vmaf": "54.63287", "config_hash": "9958b2bcc4", "git_commit": "6653a35", "videomae_b_k400_l1_lens_k16": "5.373909795336965", "videomae_b_k400_l1_lens_k32": "7.847951971760659", "videomae_b_k400_l1_lens_k64": "10.626289065084704", "videomae_b_k400_l1_lens_k128": "13.91054977635918", "videomae_b_k400_l2_lens_k16": "4.790580446716685", "videomae_b_k400_l2_lens_k32": "6.705511917496202", "videomae_b_k400_l2_lens_k64": "9.181143709298478", "videomae_b_k400_l2_lens_k128": "12.101998012874796", "videomae_b_k400_l1_pca_k64": "11.991849197437077", "videomae_b_k400_l2_pca_k64": "12.186175191228822", "videomae_b_k400_l1_full": "20.391683289407187", "videomae_b_k400_l2_full": "20.958082143240176", "video2_config_hash": "2d32247a64", "video2_git_commit": "ebe32e1", "videomae_b_ssl_l1_lens_k16": "5.22220823638928", "videomae_b_ssl_l1_lens_k32": "8.28586305933371", "videomae_b_ssl_l1_lens_k64": "11.685066500186018", "videomae_b_ssl_l1_lens_k128": "14.703533657447316", "videomae_b_ssl_l2_lens_k16": "4.883888504489623", "videomae_b_ssl_l2_lens_k32": "7.136755397725504", "videomae_b_ssl_l2_lens_k64": "9.98049884948372", "videomae_b_ssl_l2_lens_k128": "13.186436614783508", "videomae_b_ssl_l1_pca_k64": "13.071896144148551", "videomae_b_ssl_l2_pca_k64": "12.545429645294991", "videomae_b_ssl_l1_full": "21.282995973012827", "videomae_b_ssl_l2_full": "21.30794040087397", "vjepa2_l_predictor_l1_lens_k16": "8.245491862410958", "vjepa2_l_predictor_l1_lens_k32": "11.337376509123226", "vjepa2_l_predictor_l1_lens_k64": "15.579872974519422", "vjepa2_l_predictor_l1_lens_k128": "20.940355430727188", "vjepa2_l_predictor_l2_lens_k16": "8.400549533443433", "vjepa2_l_predictor_l2_lens_k32": "11.910899498488691", "vjepa2_l_predictor_l2_lens_k64": "16.670818778046527", "vjepa2_l_predictor_l2_lens_k128": "22.180009722006673", "vjepa2_l_predictor_l1_pca_k64": "19.457112662714263", "vjepa2_l_predictor_l2_pca_k64": "20.69424669596544", "vjepa2_l_predictor_l1_full": "26.777283133251046", "vjepa2_l_predictor_l2_full": "29.16930781440626", "source_line": 44 } } ], "whole_video_preferences": { "MOS": "A", "Lens": "B", "VMAF": "A" }, "crop_xywh_on_1920x1080": [ 650, 460, 384, 216 ], "crop_selection": "Manually fixed display ROI after inspecting frame 130; identical spatial window for all three videos and all three times. No score is computed from this crop.", "artifact_annotation": { "text": "A and B: texture changes", "lanes": [ "A", "B" ], "box_xywh": [ 30, 50, 140, 120 ], "frame_index": 130, "scope": "Qualitative observation in decoded pixels, not metric attribution, localisation truth or evidence of a temporal-quality mechanism." }, "original_v13_example": null } ]