原創譯文|大數據技術PK人腦：圖片、視頻分析，誰更勝一籌？

本文為燈塔大數據原創內容，歡迎個人轉載至朋友圈，其他機構轉載請在文章開頭標註：

「轉自：燈塔大數據；」

提高交通安全、改善醫療服務、提升環境效益——專家認為大數據技術在高級圖像分析和圖像識別領域潛力無限。

挪威卑爾根Uni Research公司的科學家Eirik Thorsnes表示：「計算機的高級圖像識別是一項複雜繁瑣的過程，你必須讓計算機模仿人類大腦，從大量無效信息中提取出有效信息。」

Uni Research公司大數據分析中心致力於研發大數據在研究和商業領域的應用戰略。該大數據中心還在開發高級計算機算力，模仿人腦進行複雜信息處理。

在很多方面，人腦的信息處理能力和工作方法都比電腦更加強大，但是在有些方面，電腦的表現比人腦更好。

「近幾年來我們取得了很大的發展，在圖像識別和圖像分析方面，我們的技術已經超過了人腦。電腦在觀看大量極盡相似的圖片時不會感到疲憊，而且還能夠發現人眼發現不出的細微差別。隨著我們的技術日益成熟，處理大量圖片和視頻將更加方便快捷，很多人類社會中常見的流程都能得到改進和優化，」大數據分析中心負責人Thorsnes解釋道。

識別重要目標

Thorsnes和大數據分析中心的工作夥伴預測，圖像識別和圖像分析在醫療衛生、環境監測、海底調查和衛星圖像分析等領域的重要性將日益凸顯。

大數據在圖像分析和圖像識別方面的應用，對硬體、演算法和軟體的都提出了很高的要求，同時，還要求管理者擁有卓越的能力，能夠找到最佳監測途徑。

Thorsnes 說：「在未來幾年，對這項技術的需求只會不斷增加，但是它並不是『即插即用』、能快速上手的技術。我們的研究員在處理大規模數據方面已經積累的足夠的專業知識和經驗，才能抓住最核心的應用技術。」

Uni Research計算部門的研究員開發出一套計算機系統，能夠在圖像中準確識別目標，並在圖像中發現具有重要性的對象。

人工智慧、圖像識別和機器學習方面的專家Alla Sapronova說：

「我們訓練電腦的方式和教孩子是一樣的。我向電腦輸入信號模式，並告訴它我們想要什麼樣的輸出信息。我就一直重複這個過程，直到系統開始自動識別信號模式。之後，我再給電腦展示一個新的輸入信號，比如一張電腦沒有識別過的圖片，看它是否能夠看懂。」這種機器學習技術有很大應用空間，比如，用手機相機識別笑臉。

自閉症兒童的音樂療法

這項技術的高級應用還包括醫藥領域，它可以分析身體疾病的外部信號，與臨床醫生保持溝通，檢查並報告身體狀況。

「我們已經與GAMUT合作開展了一個試點項目，分析自閉症兒童接受音樂治療的視頻錄像。通常，醫生必須花費幾個小時觀看視頻錄像，才能找到最能揭示患者精神狀態或者最能展示治療效果的鏡頭。

如果我們教電腦去識別這些畫面，那麼電腦就能去自動尋找和發現醫生想要的鏡頭，儘管到目前為止電腦還不能夠對它們進行排序。我們相信在這個領域，我們的技術擁有很大潛力。」Thorsnes說道。

在另一個項目中，研究人員將挪威卑爾根Danmarksplass十字路口的網路攝像頭作為實驗對象，教電腦識別經過該路口的車輛類型和數量。

由此可見，這項技術可應用在交通領域，幫助人們進行交通布局規劃和相關決策。此外，冬天某些時間段，Danmarksplass的空氣狀況非常差，Thorsnes認為，引入此項技術幫助優化交通布局之後，環境質量也能得到很好地改善。

Thorsnes認為圖像分析技術在改善交通安全方面具有巨大潛力，特別是在監測公路和隧道方面。計算機可以監測到不同的交通狀況，包括車輛逆行、火災、亂停的廢棄汽車、隧道里的行人等。

Thorsnes 說：「我們還能讓電腦監測主要公路邊已發生滑坡的山體，讓電腦識別什麼樣的山體變化是即將發生滑坡的徵兆。」

監測漁場的「漏網之魚」

由Klaus Johannsen 率領的Uni Research計算與大數據分析中心的團隊與Uni Research的環境部開展合作，共同監測繪製鮭魚和鱒魚在河口的運動情況。

「我們在河口的位置安裝了攝像頭，讓電腦記錄這些魚的遊動軌跡，並識別魚是野生的還是養殖的。這樣，我們就能監測是否有魚從漁場逃出來。」Thorsnes解釋道。

監測技術近幾年來取得了重大進展，其中一部分原因就是「人工智慧演算法的再發現」。

我們將行業的需求和人工智慧理念結合起來，同時，博彩業強大的計算機處理能力和複雜圖形處理系統也被我們拿來進行數據分析。

「以前，這些過程都是由人來完成的，你必須要找一個人坐在那裡盯著屏幕，看好幾個小時的醫療分析錄像或者交通路況錄像，」 Thorsnes說。

這種演算法來自「深度學習」，是具有「文藝復興式」重大意義的突破。我們的處理器已經非常先進，供我們分析的材料也越來越豐富，而且我們的計算機也已經擁有足夠的計算能力來處理更加複雜的問題，學習更「深度」的演算法。

英文原文

Using big data to analyze images, videobetter than the human brain

Improving traffic safety, better healthservices and environmental benefits -- Big Data experts see a wide range ofpossibilities for advanced image analysis and recognition technology.

"Advanced image recognition bycomputers is the result of a great deal of very demanding work. You have tomimic the way the human brain distinguishes significant from unimportantinformation," says Eirik Thorsnes at Uni Research in Bergen, Norway.

Thorsnes heads a group in the company'sCentre for Big Data Analysis focus area, which develops strategies for use ofbig data for research and commercial purposes. The Centre also works ondeveloping advanced computing power that works in the same complex way as thehuman brain.

In many areas, the human brain's fantasticcapacity and working methods will continue to outperform computers, but thereare some areas where computers can do things better.

"There has been a tremendousdevelopment in recent years, and we are now surpassing the human level in termsof image recognition and analysis. After all, computers never get tired oflooking at near-identical images and may be capable of noticing even thetiniest nuances that we humans cannot see. In addition, as it gets easier toanalyse large volumes of images and video, many processes in society can beimproved and optimised," Thorsnes explains.

Recognise which objects are importantThorsnes and his colleagues at the Centre for Big Data Analysis predict thatimage recognition and analysis will become increasingly important in areas suchas health care, environmental monitoring, seabed surveys and satellite images.

Using big data in image analysis andrecognition requires a combination of good hardware, algorithms (formulae) andsoftware, as well as people who manage to recognise the best approaches.

"The need for this kind of technologywill only increase in coming years, but it is not 'plug and play'. Ourresearchers have developed specialised knowledge about handling huge amounts ofdata, and thus how essential knowledge can be identified," says Thorsnes.

Researchers in the department Uni ResearchComputing develop computer systems that learn to recognise objects andrecognise which objects are important in the image.

Alla Sapronova is an expert in artificialintelligence, image recognition and machine learning:

"I train computers in the same way weteach children. I show the computer patterns of input signals and tell it whatI expect the output signal to be. I repeat this process until the system beginsto recognise the patterns. Then I show the computer an input signal, such as animage, that it has not seen before and test whether the system understands whatit is," Sapronova explains.

For example, on a relatively simple level,this kind of machine learning has resulted in smile recognition technology formobile phone cameras.

Autistic children undergoing music therapyMore advanced areas of application include medicine, with analysis of externalbodily signs of illness, or the detection of positive / negative situations inconsultation with a therapist.

"We have run a pilot project withGAMUT, with analysis of video footage of autistic children undergoing musictherapy. Normally, the therapist would have to spend hours reviewing thefootage to identify the exact moment that best reveals the status or progressof the patient. However, if we teach a computer what constitutes an interestingmoment, it will be able to find and select them, although to date computerscannot rank them. There is great potential for further development in asubsequent project," says Thorsnes.

In another project, the researchers used apublicly available webcam at Danmarksplass, Bergen's busiest road intersection,as a starting point to teach computers to register how many and what types ofvehicles passed through the junction during the course of the day.

This allows identification of trafficpatterns, which can then be used in planning and decision-making. In addition,at times the air quality at Danmarksplass is very poor in winter, and Thorsnesenvisages that better mapping of the traffic could also provide a basis forenvironmental improvements.

However, he believes that at the currenttime image analysis has the greatest potential in improving traffic safety,which is basically a matter of monitoring selected stretches of roads ortunnels. Computers could detect a range of different situations, including carstravelling in the wrong direction, fire, abandoned cars, people inside tunnels,etc.

"It will also be possible to getcomputers to monitor slopes susceptible to landslides along major roads, andteach the computers to recognise which changes in the landscape might imply anincreased risk of a landslide," says Thorsnes.

Monitor the incidence of escapees from fishfarms Uni Research Computing and the Centre for Big Data Analysis, headed byresearch director Klaus Johannsen, have also worked on a project mapping themovements of salmon and trout at the mouth of a river. This work was done incollaboration with another department in the company, Uni Research Environment.

"A camera was installed at the mouthof the river, and the computer was trained to record what kind of fish passed,and whether it was a wild fish or a farmed fish. In this way, we can monitorthe incidence of escapees from fish farms, among other things," saysThorsnes.

Part of the reason that detectiontechnology has made such good headway in recent years is what Thorsnes calls arediscovery of algorithms for artificial intelligence.

The industry's needs and some good oldartificial intelligence ideas found one another at the same time as massivecomputing power and sophisticated graphic processors from the gaming industrybecame available for use in analyses.

"Traditionally, these kinds ofanalyses have been carried out by people who have to sit and watch hours ofvideo footage, for example medical analysis or traffic in tunnels," saysThorsnes.

The algorithms that have had something of arenaissance come from what is now called 'deep learning', because we now haveenough computing power thanks to advanced processors and access to interestingmaterial to be able to teach more advanced and 'deeper' algorithms.

翻譯：燈塔大數據

閱讀原文了解更多詳情