williamfzc / stagesepx Goto Github PK

用于评定两个视频的相似程度（有相似的区间）

• 除了每个阶段的占用时间，从一个稳态切换到另一个稳态的耗时通常对于开发者也相当重要
• 当前的stage图是类似方波形式的，而这很容易误导人，后续改为常规折线图
• report中没有展示每个稳定阶段对应的状态图

extras中的参考值有点不明所以

这个链接的存在是否真的有必要

图表没有文字介绍

目前例子显得有点杂

改造 Reporter 与 ClassifierResult，为后续更多分类器的出现做准备

https://github.com/williamfzc/stagesepx#%E8%A7%84%E5%88%92

2019-07-25 15:51:40.061 | INFO     | stagesepx.cutter:pick_and_save:208 - pick [1, 10, 19, 28, 37] in range <VideoCutRange [1-45] ssim=0>
2019-07-25 15:51:40.061 | INFO     | stagesepx.cutter:pick_and_save:208 - pick [50, 50, 50, 50, 50] in range <VideoCutRange [50-49] ssim=0>
2019-07-25 15:51:40.061 | INFO     | stagesepx.cutter:pick_and_save:208 - pick [52, 57, 63, 69, 75] in range <VideoCutRange [52-80] ssim=0>

50-49

在分类时可以同时对每一帧进行自定义操作，例如检测视频中是否包含黑白屏。需要兼容cutter与classifier。

可能的形式：

class YourHook(CustomHook):
    # do something ...
    # and result collection ...

hook = YourHook()
cutter.add_hook(hook)
cutter.cut()
result = hook.get_result()

现在有点过于原始

目前压缩图片只支持倍率压缩，而很多时候指定尺寸的压缩才能够将一系列不同尺寸的图片标准化。

cutter会将你的视频自动切割分类出一系列的阶段，在 pick_and_save 之后通常是这样的形式：

如图，stagesepx将你的视频分成了两个阶段0与1。此时，如果你发现1中有部分图片你认为分错了（例如你认为应该属于阶段0或者应该被分成一个新阶段），你可以手动将他们移动到你认为的分类中去（例如你认为应该分类为0，你可以将这些图片移动到文件夹0下；如果你认为应该分类为新分类2，你可以新建一个文件夹2然后将这些图片移动到下面）。

在此之后利用 classifier 去分类，就可以达到误差修正的效果。你可以将该模型保存起来以便下次分析。

如果视频是拍摄所得，难免会有噪声影响。为了降低噪声对分析结果的影响，应当在分析前进行合理的预处理（例如降噪）

主要因为部分计算直接用了 ssim_list 的索引

Traceback (most recent call last):
  File "F:/stagesepx/example/cut_and_classify.py", line 9, in <module>
    stable, unstable = res.get_range()
  File "F:\stagesepx\stagesepx\cutter.py", line 228, in get_range
    self.ssim_list[end_stable_range_start_id - 1].end_time,
IndexError: list index out of range

Process finished with exit code 1

AssertionError: should load data first

已训练好的模型不需要检测有无data

maybe built with docsify

opencv原生问题，无论是图片还是视频，尽量保持所有的路径中无中文

目前的 hook 不会对原始的 frame 造成影响，也不会影响 cut 与 classify 的过程。如果该功能作为一个选项出现，hook将可以做到更多的事情（例如自定义的图片放缩、模糊与锐化等等）。

目前SVM与SSIM分类器之间的重复部分较多

例如：

• 从桌面进入应用
• 从应用返回桌面

此时，阶段1与阶段3的表现是一致的（都是桌面），如果用这样的训练结果去分析视频可能会出现不准确的情况。

出错日志如下：
2019-08-08 21:03:44.516 | DEBUG | stagesepx.cutter.cutter:convert_video_into_ssim_list:73 - ssim between 1180 & 1181: 0.8632647465179197
2019-08-08 21:03:44.678 | DEBUG | stagesepx.cutter.cutter:convert_video_into_ssim_list:72 - part 0: 0.9526492946983928
2019-08-08 21:03:44.709 | DEBUG | stagesepx.cutter.cutter:convert_video_into_ssim_list:72 - part 1: 0.9496711932140692
2019-08-08 21:03:44.737 | DEBUG | stagesepx.cutter.cutter:convert_video_into_ssim_list:72 - part 2: 0.8699858194411924
2019-08-08 21:03:44.766 | DEBUG | stagesepx.cutter.cutter:convert_video_into_ssim_list:72 - part 3: 0.956238770687535
2019-08-08 21:03:44.766 | DEBUG | stagesepx.cutter.cutter:convert_video_into_ssim_list:73 - ssim between 1181 & 1182: 0.8699858194411924
2019-08-08 21:03:44.911 | INFO | stagesepx.cutter.cutter:cut:112 - cut finished: ../boot.mp4
2019-08-08 21:03:44.930 | DEBUG | stagesepx.cutter.cut_result:get_unstable_range:70 - unstable range of [../boot.mp4]: [<VideoCutRange [1-35] ssim=[0.9181243658534559, 0.9252134306342404, 0.9050137559964935, 0.8908822591219767, 0.9148245557737175, 0.9487921674036248, 0.9490449310631784, 0.9144310146601791, 0.8778321803540652, 0.8509602102781896, 0.8338616019149875, 0.8551517699414735, 0.8825341430750818, 0.879468603985351, 0.8817909240488079, 0.9007143940379229, 0.8816951452667183, 0.8331507396734783, 0.8365107038245857, 0.8317524424284333, 0.8283431672148087, 0.8240149695572643, 0.8242853123615972, 0.8206763519559634, 0.8363518252907702, 0.8591345425489371, 0.8913256705312311, 0.8854700721409169, 0.8852091751961552, 0.8941897296160934, 0.893845546176238, 0.9111032072067128, 0.92115168086549, 0.9248221517196584]>, <VideoCutRange [36-54] ssim=[0.9420438109386463, 0.8879572430444717, 0.8904208184925245, 0.8691167000981724, 0.8995547271371701, 0.9165024337047402, 0.9078909745225309, 0.9130993683687919, 0.9153993393865681, 0.8964063241039606, 0.908572486695875, 0.8984784204612064, 0.8957711815660545, 0.8938650128215719, 0.890478744243276, 0.9082329769289119, 0.9258385267639464, 0.9224706212269499]>, <VideoCutRange [55-56] ssim=[0.947602905774728]>, <VideoCutRange [59-61] ssim=[0.938503410397641, 0.9314480675180605]>, <VideoCutRange [65-70] ssim=[0.9132356760564576, 0.8989709053962114, 0.9374417754199166, 0.9412690049439755, 0.9483129284864441]>, <VideoCutRange [71-75] ssim=[0.9125561507310234, 0.8849652489230537, 0.9132766974174759, 0.9495159705261856]>, <VideoCutRange [76-79] ssim=[0.9365373979076361, 0.9335255469226493, 0.9365607023700473]>, <VideoCutRange [82-85] ssim=[0.934846137850981, 0.9070579488291513, 0.9458062913233082]>, <VideoCutRange [87-91] ssim=[0.9115731050440241, 0.9305024839067251, 0.9314827625843571, 0.9407777027900649]>, <VideoCutRange [94-119] ssim=[0.916047299877974, 0.916395861031223, 0.8986614307258266, 0.8470687905415466, 0.8749162410711698, 0.9267084013668724, 0.9323863782676106, 0.902467677886002, 0.8801576308895069, 0.8828267262603481, 0.9363772038287094, 0.9278188632682642, 0.9004508642873242, 0.8909708647236211, 0.9075580369403887, 0.8797254519382434, 0.8596581314536917, 0.8581459222287048, 0.8939484695859322, 0.9152129598627761, 0.9275797052557968, 0.9223943000326825, 0.8929038089009321, 0.8794212849755111, 0.9232257790692727]>, <VideoCutRange [120-121] ssim=[0.9394397216950161]>, <VideoCutRange [123-136] ssim=[0.9134202718379061, 0.8929735350059819, 0.8521767988869599, 0.8626348209820505, 0.8880416572747505, 0.887639578085367, 0.8519529166718843, 0.8305632776333839, 0.8245712179159321, 0.9207230486846629, 0.896735709806936, 0.8358142525195348, 0.9071819110696226]>, <VideoCutRange [137-138] ssim=[0.9244215987082557]>, <VideoCutRange [139-153] ssim=[0.9294310180177454, 0.848478680111593, 0.8178458616011756, 0.7945456836417215, 0.8254825227478304, 0.8811278650062359, 0.8528254371737461, 0.846040360978211, 0.8413486790204538, 0.8645000680369821, 0.8995605721274038, 0.8852028256856235, 0.9096100226453765, 0.89290776040117]>, <VideoCutRange [154-168] ssim=[0.9140033554049425, 0.8699266132008844, 0.9032812608295469, 0.8703124669701822, 0.8548256615189435, 0.9033900624026308, 0.8977769428609743, 0.9353557894077562, 0.939298584336759, 0.9489880096893573, 0.8503240698366538, 0.8135336642969553, 0.8118722801034252, 0.8671877242040686]>, <VideoCutRange [169-178] ssim=[0.8900970729165435, 0.8619357871884182, 0.8901984386458397, 0.8987742001662775, 0.939710742325725, 0.9075027610172571, 0.8804409569871495, 0.8814913899332463, 0.923318861869869]>, <VideoCutRange [180-182] ssim=[0.9161308033789792, 0.9497778665341052]>, <VideoCutRange [187-192] ssim=[0.8958913072269997, 0.9038886484198713, 0.8728455336629659, 0.8879995289551622, 0.93177824899818]>, <VideoCutRange [193-202] ssim=[0.9092900811092816, 0.8627192672031541, 0.8117456523498704, 0.8457502469913444, 0.9004879707263446, 0.8778171082856986, 0.9133006619790786, 0.9162964093703343, 0.857413497404563]>, <VideoCutRange [203-214] ssim=[0.9277546582293905, 0.8717887241002067, 0.9011987398099255, 0.8761211360056683, 0.9360223153894827, 0.8864624221910704, 0.8765853943424892, 0.8751838048510959, 0.8815478700215329, 0.9164133664512916, 0.899366777946475]>, <VideoCutRange [215-220] ssim=[0.9122911059583063, 0.8752830155420894, 0.8665543861838086, 0.8700208206158795, 0.9173240381673746]>, <VideoCutRange [222-223] ssim=[0.9461697063992119]>, <VideoCutRange [227-248] ssim=[0.9374144974140738, 0.8448081508350217, 0.8545683534813803, 0.8807639045363673, 0.9124155928448431, 0.8859629618001678, 0.8660371964772295, 0.9233695315291022, 0.8849869851041774, 0.9341475331418605, 0.8750125470623177, 0.9303723170404243, 0.8902102626983182, 0.8556637098387213, 0.8905888149419433, 0.9311006854117257, 0.88500072957843, 0.8954163496667421, 0.9472951939302661, 0.9211122932306743, 0.8764018908439832]>, <VideoCutRange [249-250] ssim=[0.9476055157106805]>, <VideoCutRange [251-252] ssim=[0.9364378952921397]>, <VideoCutRange [253-256] ssim=[0.90842912924813, 0.8279834832799743, 0.8986826868956954]>, <VideoCutRange [258-264] ssim=[0.9495718710713129, 0.8801361415955551, 0.8678612398196311, 0.8629864949939425, 0.9018083371795317, 0.9403474201180007]>, <VideoCutRange [269-271] ssim=[0.9082887334587378, 0.9041532895740388]>, <VideoCutRange [273-276] ssim=[0.9217264114221823, 0.9266358229483561, 0.9298028445926105]>, <VideoCutRange [280-285] ssim=[0.9451423337751736, 0.9165878149104183, 0.9258442291920371, 0.8970289099863208, 0.9148999081314619]>, <VideoCutRange [288-300] ssim=[0.9354793839762884, 0.9288102590342234, 0.9209744284016556, 0.9220281614665765, 0.8796493839678632, 0.8980896933359289, 0.9224058019906783, 0.9128885507655224, 0.8680174714482094, 0.8747325836413995, 0.9364658774296861, 0.9376115159136487]>, <VideoCutRange [301-304] ssim=[0.9059120804090965, 0.8752413991955051, 0.9236440219018974]>, <VideoCutRange [305-327] ssim=[0.9289514427784339, 0.9213488085829686, 0.8881844522550958, 0.8651669651742009, 0.8515745315711588, 0.8508347023468081, 0.867915445398801, 0.8895093877244898, 0.887530225036962, 0.8674842297222931, 0.9456311992861018, 0.9464474613835402, 0.8967599165153687, 0.8417346723514519, 0.8187480782702747, 0.8620739192443237, 0.9284814430336455, 0.880847728261245, 0.8313400820235934, 0.8440838660478267, 0.9402011595775511, 0.8694321055474096]>, <VideoCutRange [329-337] ssim=[0.9205605870558403, 0.9299757127971737, 0.8934303859012572, 0.8917437555646135, 0.9044526973565408, 0.8559038233160416, 0.8945879077057481, 0.8916294812323253]>, <VideoCutRange [340-352] ssim=[0.9483860702480006, 0.8720659707822579, 0.9101828992228944, 0.9397991825746153, 0.9334971409750048, 0.9319786290596163, 0.8505198055454132, 0.844982967416731, 0.8778232828948758, 0.8390984705042408, 0.8274468744733717, 0.8089276614166195]>, <VideoCutRange [353-355] ssim=[0.9034374621864268, 0.894475776396278]>, <VideoCutRange [356-358] ssim=[0.9261745783809148, 0.9464140436045768]>, <VideoCutRange [360-374] ssim=[0.9045004800092085, 0.8694420610592535, 0.9202714985270322, 0.8488055974896023, 0.8331903419027629, 0.840128682250825, 0.8705451322682305, 0.8515139155987062, 0.8299057491389306, 0.824338731795251, 0.8241758925268123, 0.8544703389254413, 0.849459069179199, 0.8820111236867648]>, <VideoCutRange [378-388] ssim=[0.908964078677052, 0.9217514016490111, 0.8826313870472627, 0.8524661360946173, 0.803565351581368, 0.8401237687286791, 0.8599633063680036, 0.8491758534341126, 0.8977840465613822, 0.9358163719388369]>, <VideoCutRange [389-395] ssim=[0.9020710369612064, 0.9162835924389539, 0.9079883168024493, 0.8655188896693191, 0.8435844933921318, 0.9015380407476711]>, <VideoCutRange [397-398] ssim=[0.9304397570324507]>, <VideoCutRange [399-405] ssim=[0.889034675095657, 0.8500540017288875, 0.8521856983186764, 0.9014981335795242, 0.948138238280019, 0.9247130891072151]>, <VideoCutRange [406-408] ssim=[0.9210469016922199, 0.9202201200311455]>, <VideoCutRange [413-422] ssim=[0.9368692276930569, 0.9068012136512253, 0.8664582567701314, 0.8922037941894168, 0.8502738466424976, 0.8222292437037741, 0.8059631012974496, 0.8268044012256176, 0.9004254264832453]>, <VideoCutRange [424-429] ssim=[0.8916842000342899, 0.930982463062723, 0.9196897254772819, 0.8512136153836147, 0.905319452112295]>, <VideoCutRange [431-433] ssim=[0.9462789086431618, 0.9226733392556463]>, <VideoCutRange [435-438] ssim=[0.9439421073386318, 0.9152986449721161, 0.9276132607883923]>, <VideoCutRange [443-444] ssim=[0.9480504253275646]>, <VideoCutRange [445-452] ssim=[0.933770500136377, 0.9115190294728539, 0.9296329097863201, 0.905641948148159, 0.8871331244814475, 0.8899166801175115, 0.9376797236808186]>, <VideoCutRange [457-465] ssim=[0.9147052307353014, 0.9022742782718115, 0.928234922502605, 0.9045065348800911, 0.9373208184950615, 0.94150656014396, 0.91663293910306, 0.911092394047512]>, <VideoCutRange [466-485] ssim=[0.8873115355863911, 0.8255390582399224, 0.8389152194824806, 0.8822692693816129, 0.8665634645479956, 0.8926855574193683, 0.8629047816456885, 0.8337331958826111, 0.837740542125943, 0.8335796730621203, 0.8610546850989047, 0.8773044515582772, 0.8702353090363294, 0.8413645562564512, 0.8850898464015068, 0.8274672193148183, 0.8138275356464764, 0.8271126248683188, 0.9053474394388885]>, <VideoCutRange [487-490] ssim=[0.8835589145227666, 0.8599559856877769, 0.9240427689194268]>, <VideoCutRange [491-493] ssim=[0.9465202989042918, 0.9423305743126497]>, <VideoCutRange [495-498] ssim=[0.9356239394671415, 0.892995054649378, 0.9431273020785965]>, <VideoCutRange [501-503] ssim=[0.9239691392932715, 0.9381924625631833]>, <VideoCutRange [509-514] ssim=[0.9059703585694155, 0.873700822096657, 0.8780853378608363, 0.8879716665030188, 0.9290314024553921]>, <VideoCutRange [524-528] ssim=[0.9478127836109926, 0.8811505374397318, 0.861288037741847, 0.9373828197040298]>, <VideoCutRange [529-532] ssim=[0.9168527418555864, 0.8516199259275475, 0.8842325939494318]>, <VideoCutRange [533-544] ssim=[0.8968692079606908, 0.8946762494151175, 0.9126258141074342, 0.9349963783453182, 0.9415311444014706, 0.8934680066797938, 0.8643896432488484, 0.9054412730675832, 0.9412882249303269, 0.902969080764062, 0.9466808166331391]>, <VideoCutRange [545-547] ssim=[0.9434774629897873, 0.9338067153258096]>, <VideoCutRange [548-559] ssim=[0.9064459761915927, 0.8647310690109892, 0.935039352087305, 0.897594750078401, 0.8675410729912896, 0.8553597505740221, 0.8744997270188736, 0.8593409103206029, 0.8669258764231983, 0.8942654139613747, 0.9310592461748264]>, <VideoCutRange [566-582] ssim=[0.8826476241774445, 0.8649876556368308, 0.9170958259451594, 0.9310466546450579, 0.9423997918290264, 0.9237156483721088, 0.873460820626001, 0.8759197294128991, 0.8667082179588854, 0.8954109583191852, 0.9198430000990403, 0.9440166350425673, 0.9099168456599794, 0.8991619003243175, 0.9015269707436879, 0.9251243387963914]>, <VideoCutRange [585-586] ssim=[0.9332982613672635]>, <VideoCutRange [591-592] ssim=[0.9382452986892347]>, <VideoCutRange [593-596] ssim=[0.917221496440849, 0.8315118693223038, 0.9458815901635077]>, <VideoCutRange [600-616] ssim=[0.898805274578831, 0.8569713554279064, 0.8740773481397693, 0.9043016747091999, 0.8724067119927983, 0.8864893713622435, 0.8822620741748328, 0.8754489433259423, 0.8715629996611622, 0.9058536320964645, 0.9146580056815639, 0.8604695495555754, 0.8470467318860516, 0.8371487863891194, 0.8443209010657505, 0.9390228768361568]>, <VideoCutRange [617-618] ssim=[0.947162838699479]>, <VideoCutRange [623-625] ssim=[0.9296499581355954, 0.9019321189473875]>, <VideoCutRange [627-631] ssim=[0.9050210559962425, 0.9045797273276996, 0.9249710046157598, 0.9437905940532048]>, <VideoCutRange [635-637] ssim=[0.9187872900408469, 0.8796891141416748]>, <VideoCutRange [638-653] ssim=[0.9457327657962354, 0.8613933361585248, 0.8159989273062435, 0.8092695720421543, 0.8423951303032474, 0.8730820334519326, 0.8970818500079881, 0.91867965174877, 0.8902002888801663, 0.8952545755331156, 0.9244885596547244, 0.9203105002511481, 0.9441592538819347, 0.9277796147240458, 0.9407151769541762]>, <VideoCutRange [655-659] ssim=[0.8940794240019716, 0.8313467553727953, 0.8804681240411167, 0.9458189274471525]>, <VideoCutRange [663-666] ssim=[0.93146016671682, 0.9311122240360732, 0.9132109770501475]>, <VideoCutRange [668-678] ssim=[0.9230459464586764, 0.9191755684047082, 0.9326682612297419, 0.9370627199181669, 0.9276403821508946, 0.8618696781562908, 0.8585904826013697, 0.9354059212239474, 0.8542270768143121, 0.8642611664512977]>, <VideoCutRange [679-709] ssim=[0.8938096020607573, 0.8944075435126131, 0.8981538961118527, 0.9038952608395202, 0.869942766241637, 0.8406549703104464, 0.8307413415574989, 0.8523094883146869, 0.8870667186830079, 0.9297061050224894, 0.8920609298702413, 0.8933373180961214, 0.8614407672995864, 0.8389937047702736, 0.8324986177440564, 0.8183249998407467, 0.8266913827544792, 0.8427286969411171, 0.8697795969139793, 0.8635830870791261, 0.8887886776809801, 0.8833716024597617, 0.8375373927583256, 0.8708899612140645, 0.8687414895706435, 0.8460352895087973, 0.8386837006563117, 0.8666523616720507, 0.9275685748534643, 0.9367056704314307]>, <VideoCutRange [711-713] ssim=[0.919656348006361, 0.9438952440842463]>, <VideoCutRange [714-716] ssim=[0.9481980687107721, 0.9439539454685558]>, <VideoCutRange [719-721] ssim=[0.8596792174381301, 0.8888547691491194]>, <VideoCutRange [722-753] ssim=[0.8152956032833527, 0.8111894907369074, 0.8575681804217653, 0.8711006520089064, 0.9205106254133278, 0.8688436249053145, 0.8513564514502996, 0.8634599756820427, 0.7983889545447123, 0.7941248850060756, 0.801677012249432, 0.8020658313669452, 0.8560512061257864, 0.8783842052108646, 0.9039564832135994, 0.8892567239241517, 0.8625949529002374, 0.8487072488735241, 0.8560844709552423, 0.8740382201820481, 0.8706823340881767, 0.8630398234504868, 0.8777380022134909, 0.9115125059602605, 0.929091942710012, 0.909307366144371, 0.8549409398333172, 0.8939882994093337, 0.9326680735864133, 0.9442543775760343, 0.9479639326539914]>, <VideoCutRange [755-756] ssim=[0.9470182082136234]>, <VideoCutRange [757-758] ssim=[0.9497700687330342]>, <VideoCutRange [761-764] ssim=[0.889924211851352, 0.8570629114614624, 0.9098925359797402]>, <VideoCutRange [768-775] ssim=[0.9435810231680293, 0.9344331410371985, 0.9456859001396146, 0.9408486002996506, 0.9192998475226893, 0.9004775642184731, 0.9163017773467073]>, <VideoCutRange [776-777] ssim=[0.9357605179930808]>, <VideoCutRange [779-786] ssim=[0.9002820346720098, 0.8839124702814881, 0.8786273041357743, 0.8779116129249075, 0.9005122843835663, 0.9312293897742348, 0.9434085175639941]>, <VideoCutRange [793-794] ssim=[0.9424570551883233]>, <VideoCutRange [795-798] ssim=[0.9268508964682853, 0.9426531623858935, 0.9113684600008766]>, <VideoCutRange [799-801] ssim=[0.9466992625091026, 0.9221289049322856]>, <VideoCutRange [803-805] ssim=[0.9028150082532631, 0.9180150092092728]>, <VideoCutRange [810-819] ssim=[0.905404476323141, 0.8758223597745403, 0.9343212101718261, 0.9440689360869445, 0.8723129616838718, 0.8917536911146663, 0.8996007477769458, 0.8994171511072621, 0.9438073627806045]>, <VideoCutRange [821-823] ssim=[0.8907846565825827, 0.9338449216790483]>, <VideoCutRange [826-830] ssim=[0.8938494675055583, 0.9290850668636755, 0.8817798988544061, 0.841809183952398]>, <VideoCutRange [832-865] ssim=[0.9497781731793566, 0.9036555053479131, 0.9035729621474275, 0.913499124448536, 0.887541301266079, 0.8537740505644017, 0.8651931894701571, 0.825163376883814, 0.8062620677203294, 0.8135435592419289, 0.8503608969762342, 0.863058285733028, 0.8576904417692252, 0.8521911272561156, 0.8449405305268988, 0.865721360011569, 0.8279433892440146, 0.8069893725114491, 0.8193354906810751, 0.8728096740377417, 0.9354087768859974, 0.9419119548136581, 0.8874981072404012, 0.8775081731123501, 0.8909930076317765, 0.8850916451799296, 0.903771008773303, 0.9127169642377612, 0.882403268278562, 0.8828589816126015, 0.9197354950041999, 0.914380468302607, 0.902853149610288]>, <VideoCutRange [866-877] ssim=[0.9482916997236315, 0.8910387708003015, 0.8556988021831362, 0.8967740409854877, 0.8839524359694992, 0.8572265939448594, 0.8635260225695623, 0.8595869944882519, 0.8729166083613505, 0.91183324537747, 0.9498192418204197]>, <VideoCutRange [878-880] ssim=[0.9399682152787949, 0.9262122758282618]>, <VideoCutRange [886-890] ssim=[0.8776922347373791, 0.8272963378172297, 0.8265276891342865, 0.894294171508783]>, <VideoCutRange [891-896] ssim=[0.8913545727900511, 0.8542777986462451, 0.9024753469895676, 0.9142560740939463, 0.9308814366195813]>, <VideoCutRange [900-936] ssim=[0.9198011842814673, 0.9467728095962745, 0.8909882733630997, 0.8343960528914529, 0.8621180642655171, 0.8796668655103007, 0.8881685143690452, 0.9018050595620946, 0.8720226052600611, 0.8847048115756426, 0.9142054348782664, 0.927105738457402, 0.8755791579506641, 0.8585805424405332, 0.9034337623533307, 0.8875120572856032, 0.9183927642525744, 0.912777160842122, 0.9034441740520057, 0.9291452402256416, 0.878572360514613, 0.851799943176435, 0.8939972332724873, 0.9328538365278485, 0.8818296024737894, 0.861620421390159, 0.8461602136376278, 0.8268982009305494, 0.8209140039883873, 0.8253579660154233, 0.8581346615241656, 0.8438294838761795, 0.8461864391922872, 0.8947501797129476, 0.8975903692221429, 0.9195705082782996]>, <VideoCutRange [939-943] ssim=[0.866547752241125, 0.8546324013512915, 0.9108410735169309, 0.8966257580756022]>, <VideoCutRange [944-946] ssim=[0.9288775358545607, 0.9305635703148251]>, <VideoCutRange [949-951] ssim=[0.9204582539686503, 0.9334277301460655]>, <VideoCutRange [952-961] ssim=[0.8829721247907536, 0.8657769388216152, 0.832569030677543, 0.825498097082563, 0.8348699830250975, 0.8437044204226672, 0.913318518660412, 0.9325347348682279, 0.9386016794394554]>, <VideoCutRange [962-970] ssim=[0.932109731618047, 0.8861805335668472, 0.8315184592848889, 0.9020978723725283, 0.9327100756029562, 0.9398078184384614, 0.9221103754714078, 0.9234833870133776]>, <VideoCutRange [972-975] ssim=[0.9461337253351592, 0.9246335560913169, 0.9303903487516243]>, <VideoCutRange [976-979] ssim=[0.8796815355172819, 0.8574075927593574, 0.9196643635997797]>, <VideoCutRange [980-991] ssim=[0.9201430085983604, 0.8948770086937307, 0.8548221151146561, 0.8231704913512738, 0.8339846795309062, 0.8303058359112415, 0.8828150884754162, 0.8873142497287798, 0.8851241189407923, 0.8792639943003658, 0.8816194383032732]>, <VideoCutRange [993-996] ssim=[0.8481433779606206, 0.82819544381876, 0.8780530999077077]>, <VideoCutRange [999-1036] ssim=[0.9021088338878132, 0.8439345830034864, 0.8122104864552475, 0.8012777522808433, 0.8213870691174333, 0.8902908471973925, 0.9054097596470879, 0.9069159205766527, 0.928481723890282, 0.8428272901720049, 0.8269271389835722, 0.8323064105906846, 0.8137088150262268, 0.8163369032726366, 0.8373540463033557, 0.832022370834878, 0.834964047640812, 0.8327453786205329, 0.8393628687921669, 0.8446965507649976, 0.8462924705475026, 0.8441331323213942, 0.8414834091438749, 0.8648394338099822, 0.8627311532604859, 0.8506416379734875, 0.856733774690456, 0.8809164686775155, 0.9026636017718463, 0.8978973841189188, 0.843222211344476, 0.8970973230431188, 0.9135049173878326, 0.8794307463060904, 0.8869139123158498, 0.8999378610983279, 0.9267418786378968]>, <VideoCutRange [1039-1046] ssim=[0.8965104705467255, 0.866210499275293, 0.8507370179523148, 0.9029850175370409, 0.9412207039525108, 0.9306024223983564, 0.9205103437449235]>, <VideoCutRange [1048-1063] ssim=[0.8850476301947425, 0.867286830453785, 0.8533341823169079, 0.8649365461688545, 0.8589542734450348, 0.8507962092694065, 0.8355429705938386, 0.832151686801738, 0.8351945837543192, 0.8463750044205267, 0.8924960956702307, 0.8458968107517036, 0.853862234393965, 0.8926972587359329, 0.9404369824822219]>, <VideoCutRange [1064-1066] ssim=[0.9337430051848495, 0.9485445472169387]>, <VideoCutRange [1067-1074] ssim=[0.904777770611344, 0.9198752081395798, 0.8999945744862357, 0.911333647499567, 0.8748087539134144, 0.9268799769370024, 0.9316383808740144]>, <VideoCutRange [1079-1091] ssim=[0.9269957477590604, 0.8669993787936, 0.9117628469729593, 0.8841693250296185, 0.8522786907308836, 0.8621037553914498, 0.8554140811381713, 0.8435815702471774, 0.8294220037944929, 0.8595023289080738, 0.9104647893708415, 0.939341564446912]>, <VideoCutRange [1092-1093] ssim=[0.9335624336444992]>, <VideoCutRange [1094-1098] ssim=[0.9193405879283291, 0.8736799989076443, 0.9378308554568376, 0.9101265227882159]>, <VideoCutRange [1100-1115] ssim=[0.9349324111271454, 0.9038841192948003, 0.8938413308169831, 0.903395017083237, 0.9229255144472905, 0.8977293734899444, 0.8740679635826715, 0.852301953993077, 0.8542147727522732, 0.8971661255620301, 0.8991362225832089, 0.8868781799192135, 0.9015302541505905, 0.9359133414929117, 0.9470193994920513]>, <VideoCutRange [1119-1128] ssim=[0.9095540454031426, 0.8212552332385732, 0.5631660269333111, 0.8090959223618649, 0.8565985758265745, 0.885134091174482, 0.9267166874126134, 0.9381303904897996, 0.9479077893356675]>, <VideoCutRange [1130-1131] ssim=[0.9447442855287157]>, <VideoCutRange [1132-1148] ssim=[0.9133557367706386, 0.8821354743016915, 0.8768338598781673, 0.8806029364224163, 0.921979558899471, 0.8801706529016198, 0.8442989685368619, 0.833165679111592, 0.8770430624423303, 0.8750905641659973, 0.8210279866799485, 0.8176366527254532, 0.8260942183970337, 0.9039326642629166, 0.9419427270830297, 0.9380884809419646]>, <VideoCutRange [1149-1158] ssim=[0.7995059454891394, 0.7618970442740022, 0.7983324575595944, 0.826169305834383, 0.8260546703898662, 0.8046866316071338, 0.7990330706885145, 0.8820629471962438, 0.9052113868454351]>, <VideoCutRange [1161-1162] ssim=[0.824272589676426]>, <VideoCutRange [1163-1182] ssim=[0.8581314334881646, 0.8437381874267059, 0.8438084961304544, 0.9226115522276603, 0.8363695590725925, 0.9061433549206566, 0.9054492066154005, 0.8928011202283437, 0.9204247232735968, 0.866334985826736, 0.7938396223253756, 0.8091568714394337, 0.8198893745575723, 0.8478278434858042, 0.856499193832517, 0.9120901761760489, 0.8827618675332891, 0.8632647465179197, 0.8699858194411924]>]
Traceback (most recent call last):
File "D:/PycharmProjects/stagesepx/example/cut_and_classify.py", line 9, in
stable, unstable = res.get_range()
File "C:\Python37\lib\site-packages\stagesepx\cutter\cut_result.py", line 114, in get_range
self.get_target_range_by_id(first_stable_range_end_id - 1).start_time,
File "C:\Python37\lib\site-packages\stagesepx\cutter\cut_result.py", line 23, in get_target_range_by_id
raise RuntimeError(f'frame {frame_id} not found in video')
RuntimeError: frame -1 not found in video

for further usage, i choose imageai as AI backend, which was built on the top of keras and tensorflow.

stagesepx have no plan of offering an built-in deep learning toolbox (too much dependenices it needs). stagesepx is a lightweight tool for analysing videos only and supports running on the most of platforms.

but we will offer a workflow (example) to show how to make it work with popular AI frameworks.

Traceback (most recent call last): File "D:/pythonworkpace/stagesepx/example/compare_videos.py", line 13, in <module> stable, _ = res.get_range(limit=3) File "D:\pythonworkpace\stagesepx\stagesepx\cutter.py", line 197, in get_range unstable_range_list = self.get_unstable_range(limit, **kwargs) File "D:\pythonworkpace\stagesepx\stagesepx\cutter.py", line 168, in get_unstable_range if change_range_list[-1].start > merged_change_range_list[-1].end: IndexError: list index out of range
拉去的是最新代码

2019-08-15 22:11:48.006 | INFO     | stagesepx.cutter.cut_result:pick_and_save:274 - pick [0] in range <VideoCutRange [0-18] ssim=[1.0]>
2019-08-15 22:11:48.006 | DEBUG    | stagesepx.cutter.cut_range:pick:102 - pick 1 frames from 20(0.8) to 21(0.8) on video ../demo.mp4
2019-08-15 22:11:48.051 | INFO     | stagesepx.cutter.cut_result:pick_and_save:274 - pick [20] in range <VideoCutRange [20-21] ssim=[1.0]>
2019-08-15 22:11:48.051 | DEBUG    | stagesepx.cutter.cut_range:pick:102 - pick 1 frames from 30(1.2) to 30(1.2) on video ../demo.mp4

frame 20 and 21 have the same timestamps.

目前 classifier 只能通过 切割好的图片 进行分类，而不能直接利用 cutter 的返回结果。
在不需要中间结果的情况下，这会造成一定的IO浪费。

目前在评估stagesepx效果时，对于阶段划分是否准确的评估是比较麻烦的。

一些关键API的注释补全

具备切割器之后，stagesep2的性能问题可以得到解决。

SSIM原生对高斯模糊不敏感，而目前非常多的应用带有blur类型（类似毛玻璃）的动画效果。

视频比较功能能够让stagesepx真正被用于功能校验上

对于稳态与非稳态，limit参数的价值完全是不同的。而目前的做法是一个limit参数会同时影响到他们两者的过滤，需要将他们区分开。

for each in unstable:
    r.add_thumbnail(
        f'{each.start}({each.start_time}) - {each.end}({each.end_time})',
        res.thumbnail(each))

r.draw(
    classify_result,
    cut_result=res,
)

目前 thumbnail 与 ssim+psnr趋势图 都是分别可选的，而他们在debug中举足轻重。
保留一处开关即可。

目前compress方法带了一个简单的降噪流程，这应该以hook形式实现以达到便于拼接的目的。

非常值得注意的是，分析结果的准确性与如何录制视频有较大关联性。以android为例，通常使用adb shell screenrecord 录制出来的视频，fps是不稳定的。

The frame rate is variable, not fixed. Every time the screen is updated, one frame is recorded. If the screen is not updated, no frame is recorded. Therefore there is no setting for the frame rate, because it's determined by how quickly the system updates the screen.

而这种不稳定很可能带来非常巨大的误差，绝对不应该被忽视。

如上述例子，同一个场景下，不同的录制方法导致的误差达到了3s之多。

目前推荐两种视频录制方式：

• 软件录制+ffmpeg（必备项，具体可参考 https://github.com/williamfzc/sepmachine 的实现）
• 硬件（外置摄像机）

对于已有视频，请参考：

ffmpeg -i input.mp4 -r 60 output.mp4

目前 VideoCutRange中的属性有：

self.video_path = video_path
self.start = start
self.end = end
self.ssim = ssim

时间戳

事实上，视频与帧位置都已经确定了，对应的时间戳其实也已经确定，完全可以在其中加入起始与结束的时间戳。

SSIM列表

VideoCutRange里应该是SSIM列表，而不是一个已经评估完的SSIM具体值

目前stagesepx中存在部分重复代码是因为重复从视频中获取frame对应的时间戳与id等，计划用 VideoFrame 替代。

Traceback (most recent call last):
File "test01.py", line 10, in
data_home = res.pick_and_save(stable, 3)
File "/python/lib/python3.6/site-packages/stagesepx/cutter.py", line 186, in pick_and_save
each_frame = toolbox.get_frame(cap, each_frame_id - 1)
File "/python/lib/python3.6/site-packages/stagesepx/toolbox.py", line 48, in get_frame
assert ret
AssertionError

2019-08-22 15:16:27.751 | DEBUG    | stagesepx.cutter.cut_range:pick:102 - pick 5 frames from 50(0.8833333333333334) to 112(1.8833333333333333) on video ../demo.mp4
2019-08-22 15:16:28.407 | INFO     | stagesepx.cutter.cut_result:pick_and_save:284 - pick [50, 62, 75, 87, 100] in range <VideoCutRange [50-112] ssim=[1.0]>

首帧始终会被采样到。而对于稳定阶段而言，选择首帧是有风险的（可能它并没有彻底变化完毕，只是低于阈值）。尤其当采样张数较少时，它的权重会更大，从而导致效果不准确。

帧级别的操作全部以hook形式实现，具备更高的灵活性

AI分类器需要大量的训练集，而这不可能来自同一个视频。

在实际工作中，很多时候需要对report进行定制，例如插入更多相关信息。

目前当 cutter 与 classifier 分开使用时，仍旧有部分功能是不支持的（例如将变化过程置为-1，而这非常影响SVM的分析效果）。

理想的目标是，无论分开与否，两者应该能够表现一样的功能。

https://stackoverflow.com/questions/189943/how-can-i-quantify-difference-between-two-images

为了提高该库的稳定性与适用性，我们每次发布前都会在内部视频测试集上运行以确保效果正常。但测试集的数量有限，我们很难保证所有场景的使用效果。

欢迎所有的开发者提交你觉得目前 stagesepx 支持得不够好的测试视频（当然，尽量不要涉密，可以是你的使用场景也可以是你认为分析不准确的场景），把这个工具做得更好。

视频请不要超过半分钟，短视频、MP4格式为佳。

请以邮件形式发送到：[email protected]
或以附件形式贴到该issue中。

快照是直接由帧紧凑拼接而成，中间无间隔。而当背景一致时，人眼阅读比较困难。