AI攝影助手是繼AI攝影大師之后在AI攝影領(lǐng)域的另一大突破方向。與AI攝影大師有著根本不同的一點(diǎn)是：AI攝影助手將不僅僅進(jìn)行基于場(chǎng)景識(shí)別所獲取的信息進(jìn)行識(shí)別，而是采納更多信息，比如地理位置、時(shí)間、天氣等，同時(shí)還會(huì)接收來(lái)自拍攝者的選擇，從而得出更加精細(xì)化的調(diào)整方案。

功能：

1、基于語(yǔ)義分析的問(wèn)題機(jī)制，以華為影像實(shí)驗(yàn)室舉例。將參賽照片進(jìn)行更加細(xì)節(jié)的語(yǔ)義分析。包括拍攝時(shí)的時(shí)間、地點(diǎn)、場(chǎng)景中的元素，以及針對(duì)照片進(jìn)行的修圖操作。其中，時(shí)間地點(diǎn)元素這些內(nèi)容為判斷元素，每當(dāng)當(dāng)前場(chǎng)景與庫(kù)中的照片有較大符合度時(shí)，便可以將建議推送給使用者，以縮略圖的形式將構(gòu)圖提供給使用者，并且將其修圖操作實(shí)時(shí)呈現(xiàn)在拍攝界面中。同時(shí)，如果使用者可以提供已經(jīng)進(jìn)行過(guò)語(yǔ)義分析的照片，AI攝影助手可以將其信息讀取后為使用者提供近似的效果，同時(shí)會(huì)指出當(dāng)前場(chǎng)景同給定照片的不同之處。

1.??problem mechanism Based on the semantic analysis. take Huawei Imaging Laboratory as an example. They made ?detailed semantic analysis on the photos of the huawei new image competition. It includes ?time and place of shooting, the elements in the scene, and the photo editing operation. Among them, the contents of time and place elements are judgment elements. Whenever the current scene has a large consistency with the photos in the library, you can push suggestions to the user, provide it to the user in the form of thumbnails, and present its revision operation in the shooting interface at real time. At the same time, if users can provide photos that have been semantically analyzed, AI photography assistant can provide users with similar effects after reading their information, and point out the differences between the current scene and the given photos.

2、無(wú)場(chǎng)景的推薦機(jī)制。針對(duì)只有大致拍攝方向的人群，在無(wú)法分析場(chǎng)景的情況下，可以采取更為模糊的推薦機(jī)制。例如：白天、運(yùn)動(dòng)、高速這些特征詞輸入后，應(yīng)該將其判斷為運(yùn)動(dòng)比賽類場(chǎng)景，進(jìn)行加快快門速度,縮小光圈的操作；當(dāng)輸入拍人、人像、給女朋友拍照等特征詞時(shí)，應(yīng)該判斷為人像場(chǎng)景，自動(dòng)將光圈放大。

?

2. None-scenario recommendation mechanism. For people who only have a general shooting plan, a more vague recommendation mechanism can be adopted when the scene cannot be seen. For example, taking?daytime, sports and high-speed feature words as?input,AI image assistant should judge it?as sport competition scene, and the shutter speed should be increased and the aperture should be reduced; When you enter the key?words such as taking a portrait, or taking a picture of your girlfriend, AI image assistant should judge it as a portrait scene and automatically enlarge the aperture.

?

3、基于輸入的教學(xué)機(jī)制。在推薦機(jī)制之上，當(dāng)使用者對(duì)于即將做出的調(diào)整表示不理解之時(shí)，可以對(duì)使用者播放預(yù)設(shè)的視頻，讓使用者得到提前學(xué)習(xí)。

3. Input based teaching mechanism. On the recommendation mechanism, once?the user does not understand the adjustment to be made, the user can play a?pre-set video so that the user can learn in advance.

4、基于給定圖片的輔助機(jī)制。例如使用者在社交平臺(tái)上看到了一種符合審美的作品，想要進(jìn)行模仿時(shí)，該照片的作者先向AI攝影助手提供相對(duì)應(yīng)的拍攝參數(shù)，然后AI攝影助手會(huì)根據(jù)所給定的參數(shù)引導(dǎo)使用者進(jìn)行拍攝。

4. An auxiliary mechanism based on a given picture. For example, when users see an aesthetic work on the social platform and want to imitate it, the author of the photo first provides the AI photography assistant with the corresponding shooting parameters, and then the AI photography assistant will guide users to shoot according to the given parameters.

5、基于隨機(jī)化與給定照片的評(píng)價(jià)機(jī)制。在拍攝完畢后，AI攝影助手會(huì)搜尋與其相似度較高的照片，通過(guò)分析二者參數(shù)的不同，將超過(guò)閾值的不同以隨機(jī)化的策略向使用者指出。

5. Evaluation mechanism based on randomization and given photos. After shooting, the AI photography assistant will search for photos with high similarity to them, and point out the difference exceeding the threshold value to the user with a randomized strategy by analyzing the difference between the two parameters.

6、基于光源標(biāo)注的圖像重建技術(shù)。以往，想要將一張照片的元素轉(zhuǎn)換到另一張照片上時(shí)，要考慮的最大因素就是光源問(wèn)題。通過(guò)AI光源分析，或是AI與人工標(biāo)注相結(jié)合的混合光源分析，將兩張照片的光源情況進(jìn)行比對(duì)，從而在元素移動(dòng)后更加符合環(huán)境，減少不自然感。

6. Image reconstruction technology based on light source annotation. In the past, when you want to convert the elements from?one photo to another, the biggest Challenge?to consider is the light source. Through the AI light source analysis, or the mixed light source analysis combining AI and manual annotation, the light source of the two photos is compared, so that the elements move more Correctly with the environment and reduce artificiality?in the scene.

7、自定義參數(shù)的存儲(chǔ)與讀取系統(tǒng)。AI攝影助手可以開啟專業(yè)模式，讓使用者自定義參數(shù)，然后將參數(shù)保存為一個(gè)預(yù)設(shè)，然后在讀取時(shí)，可以控制這一預(yù)設(shè)所使用的條件：例如xx場(chǎng)景下使用，今天一整天要使用這個(gè)預(yù)設(shè)，室外使用這一套參數(shù)配置等等，為專業(yè)人士提供視覺(jué)化的編程方案，讓使用者能夠最大程度提升效率。

7. Customized parameter storage and reading system. The AI image?assistant can turn to?professional mode, letting?users to?customize the parameters, and save the parameters as a pre-set. Then, when reading, users?can control the conditions used by this pre-set: for example, in this?scene, this preset is used all day today,this set of parameter configurations are used outdoors，etc. It provides professionals with a visual programming scheme, so that users can maximize?their?efficiency.

8、頻數(shù)智能感知系統(tǒng)。AI攝影助手將自動(dòng)分析并篩選出拍攝頻數(shù)最高的人物，詢問(wèn)并將其添加為常用人物。進(jìn)行對(duì)焦時(shí)會(huì)優(yōu)先選擇該人物，并且根據(jù)精準(zhǔn)的圖像分割系統(tǒng)檢測(cè)人物的妝容、衣著，將不同的打扮穿搭風(fēng)格分別記錄下來(lái)，分別應(yīng)用不同的效果。

8. intelligent?Frequency sensing system. AI?image?assistant will automatically analyze and screen out the most frequently photographed people, ask them and add them as close?mate. When focusing, that?person will be selected preferentially, and the makeup and clothing of the person will be detected according to the accurate image segmentation system. Different dressing styles will be recorded and different effects will be applied.

9、簡(jiǎn)單模式與高級(jí)模式。簡(jiǎn)單模式下，AI攝影助手會(huì)更多傾向于引導(dǎo)使用者，參數(shù)仍由自己控制，但是在此之上的統(tǒng)合功能皆由使用者決定最終效果，同時(shí)，運(yùn)用語(yǔ)義分析系統(tǒng)，將使用者抽象的想法轉(zhuǎn)化為可調(diào)整的參數(shù)，讓使用者的想法能夠得到實(shí)現(xiàn)。高級(jí)模式下，AI攝影助手的引導(dǎo)部分將會(huì)減弱，各個(gè)功能保持不變，而且開放具體參數(shù)控制給使用者，讓使用者針對(duì)不同條件快速調(diào)整參數(shù)，提高拍攝效率。

9. Simple mode and advanced mode. In the simple mode, AI image?assistant tend to guide users more, and the parameters are?controlled by AI, but the integration functions above are all determined by users. At the same time, with semantic analysis system，AI image?assistant?can convert users' abstract ideas into adjustable parameters, so users' ideas can be realized?correctly. In the advanced mode, the guidance part of the AI photography assistant will be weakened, and each function will remain unchanged. In addition, specific parameter control will be opened to users, allowing them to quickly adjust parameters according to different conditions to improve efficiency.

需求：

1、經(jīng)過(guò)語(yǔ)義分析的圖像格式。當(dāng)前的主流電子照片格式基本都支持通過(guò)EXif對(duì)細(xì)節(jié)進(jìn)行記錄，但是要達(dá)到更加精細(xì)的語(yǔ)義分析，需要對(duì)EXif版本進(jìn)行更新（或者更改），除去基本的光圈、快門、ISO，位置、時(shí)間、焦距、白平衡等信息以外，還需要增加俯仰角、傾斜角、方位這些傳感器可以感知的數(shù)據(jù)，以及場(chǎng)景元素這些語(yǔ)義分析信息（這一部分是提前存儲(chǔ)好信息，節(jié)省實(shí)時(shí)分析所消耗的資源），以及更加抽象的數(shù)據(jù)，例如“是否為室外”、“光源位置以及光源屬性”，最后還有對(duì)照片所進(jìn)行的修圖步驟。

2、擬人形態(tài)的AI形象。AI攝影助手的目的就是通過(guò)引入人類決策時(shí)會(huì)進(jìn)行的選擇、判斷、決定之類的操作來(lái)達(dá)到讓AI更像人類的目的，所以一個(gè)擬人化的形象可以適當(dāng)提高使用者的沉浸感。

3、與語(yǔ)義分析、圖像重建等先前已有AI技術(shù)的融合。AI攝影助手并非獨(dú)樹一幟，而是對(duì)前者的繼承，所以也要求先前的技術(shù)作為新系統(tǒng)的一部分所保留下來(lái)。

4、聯(lián)網(wǎng)能力。現(xiàn)實(shí)證明，將龐大的輸入圖像庫(kù)直接塞到每個(gè)使用者的設(shè)備中基本被證明為不可能的，即使將圖像進(jìn)行有損壓縮也是件難事，所以在一些功能使用的時(shí)候，必須要通過(guò)網(wǎng)絡(luò)與數(shù)據(jù)庫(kù)進(jìn)行通信才能達(dá)到正常的效果。

5、隱私問(wèn)題。如此繁多的信息記錄必然會(huì)帶來(lái)對(duì)個(gè)人隱私問(wèn)題的挑戰(zhàn)，如何解決這個(gè)問(wèn)題是AI攝影助手能否進(jìn)一步發(fā)展的關(guān)鍵。

（純屬虛構(gòu)）