The service is used to verify the similarity between two faces. It can be used in the internet bank, insurance and entrance guard. In the LFW dataset, our real time accuracy is over 99%.
The service includes detection of gender, age, race, expression and face landmark. It can be used in many areas, such as photo refinement, avatar and predestination by photos.
The service is to recognize 20000 VIPs from politician, actors, sportsman or other focus persons. It can be used in automatic person detection from TV videos, suspect monitoring. The top-1 accuracy is 91%.
The service is to recognize objects or scene in photos. 2000 categories are predefined, such as clothes, home furnishing, animal, violence and pornography. It can be used in shopping, automatic album analysis and internet security, etc.
This service is used to detect whether there are pornographic contents in images.
The service is to recommend VIPs to users in news and serial videos by face detection, clustering, tracking and recognition algorithms.
effective, steady and massive recognition engine from the cloud server.
Off-line SDK: provide SDK grant authorization for special enterprises with very large requirements.
Custom private cloud: recognition servers will be delivered to authorized enterprises.
The demo includes three steps: (1)press the following white cross button and upload two images, (2)select one face square region and the square will be change to red, (3)after press the COMPARE button, the similarity will be shown. More details
After you press the following white cross button and upload one image from the left column, face attributes will be shown in the right column. More details
The demo includes three steps: (1)press the following white cross button and upload one image from the left column, (2) select one face square region and the square will be change to red, (3)after press the RECOGNIZE button, the similarity will be shown. More details
Press the following white cross button and upload one image from the left column. The detected concepts will be shown. details
Press the following white cross button and upload one image from the left column. Then the result will be shown.
If the QR code doesn't work, try downloading our demo in following ways:
Android: search and install Faceall Liveness Detection in qqdownloader.
Apple: search and install Faceall Liveness Detection in AppStore.
Beijing FaceAll Co. is National High-Tech Enterprise, Beijing Software & High-Tech Enterprise, authorized by Beijing Government, China. It was founded at 2015, and is Angel invested by Huami(Xiaomi Bracelet) Co. in China at 2016, and is Pre-A invested by Qingdao Dingyuan Co. at 2017. We focus on the algorithmic core technology innovation of face recognition, image recognition and video retrieval by machine learning and especially deep learning algorithms. Many of our key members have working experiences at top 500 Enterprises in the world. Our vision is to develop the best face recognition and image recognition software services for related enterprises.
We have achieved many rewards in world-level face recognitoin, face detectoin, multimedia retrieval, image recognition and face recognition benchmark challenges. The online API service and offline sdk have been developed for individuals and businesses, including face detection, face verification (1:1), face identification (1:N)，face landmark detection(5 points, 27 points, 68 points, 194 points), celebrity recognition, face attribute analysis(genders, age, races) , face tracking , face liveness detection and object/scene recognition.
We present world-level real-time software services on:
1. Face Detection In world-level face detection benchmark at FDDB, we have achieved the accuracy at 98.74%. , which means good performance in world-level.
2. Face verification (1:1) In face verification benchmark at LFW, we have achieved the accuracy at 99.67%, which means good performance in world-level, when Google achieves the accuracy at 99.65 in the same benchmark corpus
3. Face identification (1:N) In Megaface 2015 face identification benchmark, when searching 1 face from N=1million faces, our engine has reached 76.6% accuracy, when Google has reached the 70% accuracy in the same evaluation set .
4. In the aspect of 68 face landmark points detection, we achieved an error rate of 5.31% in the 300 Faces In-the-Wild Challenge (300-W), better than the best performance of 6.31% in the literature.
5. We have over 20000 celebrities face corpus, so we could recognize much more famous person's faces in the world.
6. On Oct 2017 , we participate ICCV2017, Venice, Italy , VOT(visual object tracking challenge)2017, which is the top level object tracking challenge in the world. We won top 2 among 60 submissions from 38 top level research units in the world .
Our image recognition engine could recognize over thousands of objects/fields. It could also do pornographic image detection, and terroristic picture (such as ISIS, bomb, gun, terrorist faces).
We do not only make effort on theoretical face recognition and image recognition technology benchmark, but also work on the recognition to the real field’s faces and images, to make our engine the best face recognition and image recognition engine in the world.
We have world-level business cooperator of telecommunication operator , telecommunication manufacturer, internet service provider , in the Service at automatic TV video analysis, entrance guard, internet payment, checking passports and pornographic photo detection , advertisement recommendation etc.
We present face recognition service, image recognition service, pornographic image detection service ， and we also present offline sdk support for your own private clould services.