H.V.A.S.

Hora Video Analytic System

Try all the models at app.horavision.ai by making a free account. See here for how to use the platform or the API.

Face and person recognition models in Python using neural network architectures including Yolo and EfficientNet, as well as various features such as gender, age, head and body pose, gaze estimation, landmarks, mask usage and clothing. These models can detect more faces quicker than Google Cloud, Amazon Rekognition, and Azure Vision.

Face Recognition models

• Face Detection
  • Gender Detection
  • Emotion Recognition
  • Age Estimation
  • Head Orientation Estimation
  • Gaze Estimation
  • Facial Landmarks Detection
  • More Facial Landmarks Detection
  • Face Mask Usage
  • Face Recognition

Performance

Model	Arch	Complexity (GFLOPs)	Size (Mp)	AP @ [IoU=0.50:0.95] (%)
Face Detection	Yolo v5	29.4	20.0	98.7

Model	Arch	Complexity (GFLOPs)	Size (Mp)	AVG Top-1 (%)
Gender Detection	EfficientNet	3.76	14.14	98.3
Emotion Recognition	EfficientNet	4.22	19.33	72.3
Face Mask Usage	EfficientNet	2.85	8.41	99.4

Model	Arch	Complexity (GFLOPs)	Size (Mp)	MAE / MNE
Age Estimation	MobileNet	0.76	4.14	5.5
Head Orientation Estimation	SimpleCNN	0.210	3.82	7.1
Gaze Estimation	SimpleCNN	0.268	3.77	6.2
Facial Landmarks Detection	SimpleCNN	0.056	1.76	0.11
More Facial Landmarks Detection	SimpleCNN	0.142	4.25	0.56

Model	Arch	Complexity (GFLOPs)	Size (Mp)	LFW
Face Recognition	MobileNet V2	5.88	11.07	0.9832

Demo:

Person Recognition models

• Person Detection
  • Person Clothing Recognition
  • Pose Estimation
  • Person Recognition

Performance

Model	Arch	Complexity (GFLOPs)	Size (Mp)	AP @ [IoU=0.50:0.95] (%)
Person Detection	EfficientDet	22.4	16.4	98.1

Model	Arch	Complexity (GFLOPs)	Size (Mp)	AVG Top-1 (%)
Person Clothing Recognition	EfficientNet	13.66	51.64	90.5

Model	Arch	Complexity (GFLOPs)	Size (Mp)	AP (%)
Pose Estimation	MobileNet V2	50.25	23.16	79.0

Model	Arch	Complexity (GFLOPs)	Size (Mp)	mAP (%)
Person Recognition	MobileNet V2	4.01	5.10	96.2

Demo:

Comparison with Google Cloud, AWS Rekognition, and Azure Vision

Google Cloud

Face count: 51

AWS Rekognition

Face count: 92

Azure Vision

Face count: 103

HoraVision

Face count: 176