Google Vision API v/s Microsoft Vision API

This week, I have been working on image recognition and the two popular services which I tried Google Vision API and Microsoft Vision API.

Both of them are promising and really make our life easy with a new way to think.

Feature comparison

Feature Google Vision Microsoft Vision Objects Detection Yes Yes OCR Yes Yes Facial Detection Yes Yes Landmark Detection Yes No Logo/Brand/Person Detection Yes Yes Inappropriate content detection Yes Yes

Objects detection

For object detection both the API gave pretty decent results, I tried a fruit salad image and another one from a beach.

2016-05-12

Source: videezy.com

Here are the results I got

Google Microsoft Label Annotations:

vacation — Score:0.95443827

swimming pool — Score:0.67789888
 People detected: 1 Image Details:(

water,outdoor,person,woman,beach,

holding,young,standing,sitting,smiling,boat,surfing,

wearing,white,body,girl,man,board,ocean)

Clearly Microsoft Vision API object recognition is more detailed Over Google Vision API.

OCR

2016-05-12 (1)

Google Microsoft BAZAAR,Apple,200 Hi,iPhone SE,Mi5,Price,Gear,VR,subscriber copy not for sale,A99 Media,Publication,YOUR TECHNOLOGY NAVIGATOR,WW W. DIGIT,IN I VOL.16 ISSUE 05,THE,Predator,15,CHEA,WI,REPEATER,REPAIR,TECHNOLO,EXPI,GRAVITATION,WAY,DISCO,EDUCATI,STARTU,ENGINEERED SUPER your

OCR of Google vision API is much much better than Microsoft Vision API. Microsoft worked little better in B/W image scans but Google Vision was amazing. So if you are looking more for OCR go for Google vision API. So far there is a tie between both 1/1.

Facial Detection

2016-05-12 (2)

Source :http://www.provincia.carboniaiglesias.it/

Google Microsoft 10 People detected full of Joy. 15 faces detected, out which 14 were Happy and 1 neutral

Both of them were pretty accurate in Facial detection, Microsoft was little better with more accuracy and detection.

Microsoft also gives you a lot of face matching , grouping similar faces and face search features with its Face API. This a pretty cool in a lot of use cases. Google

Logo/Brand/People/landmark Detection

quantico-saison-1-abc-episode-1-episode-2

Source: http://www.melty.fr/

Logo/Brand — Google was able to recognize most of the brands but Microsoft was not heavy on this one.

People — Microsoft has a celebrity detection feature which is really cool, Google visio doesn’t support this so far. It was able to detect most celebrities like the famous girl from Quantico.

Landmark — Google vision API is really good in this as compared to Microsoft vision.

Inappropriate Content Detection

Kissing images — Both of the API’s were ok

Adult/Nude content — Both the API’s were able to identify and label images accurately. Google was little better than Microsoft in identifying images like operations different from adult images.

This is a pretty usecase powerful for all product which rely on user generated content.

Overall Google vision API is giving better results for most image recognition use cases and Microsoft is way ahead in Face recognition and emotions with its Face API and Emotions API


Originally published at Next Gen Technology Insider.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.