Scene Classification

TetrasMobile
Smart Video

Supported by the world's leading AI visual perception technology, Tetras intelligent video solution focuses on the comprehensive information recognition of faces, human bodies, objects, spatial scenes, and video content. Based on the application scenarios of customers in different industries, we provide products and services in both images and videos, such as facial feature verification, face clustering, facial attribute analysis, structured analysis of scene content, and video highlight generation. At the same time, we support the large-scale deployment of mobile devices and the cloud, which provides efficient and accurate AI perception for multitudinous scenarios, such as intelligent photo album management, best moments capture, intelligent video editing, and media content searching.

PRODUCT INTRODUCTION

Product Introduction

Tetras.AI's scene classification, through expansive image data mining of life scenes, can accurately undersatand picture and video content information and realize intelligent classification, helping users to better manage mobile phone photo albums. The scene classification algorithm supports mainstream platforms such as Android and iOS operating systems, as well as cloud services, with accuracy and recall rates surpassing 90% and 70%, respectively. At the same time, the algorithm supports the classification of 92 categories of picture labels, 46 categories of video labels and thousands of sub-categories of labels, and supports single-label and multi-label output. Users can quickly expand sub-category labels or formulate personalized labels according to actual usage requirements to improve the photo album classification management experience.

Supported Platform

Supports Android and Linux platforms; and the CPU version supports a full range of Qualcomm and MTK product models.

PRODUCT ADVANTAGES

Product Advantages

1. Scene Classification products can be customized according to requirements, process massive data, rapidly expand the number of labels, and establish a hierarchical structure for label content classification and relational mapping.

2. The hierarchical level can be selected according to the requirements and needs. Single-label and multi-label output are supported, which line up with the actual user's requirements. Without the platform, frame rate, and resolution restrictions, it can achieve millisecond-level recognition.

KEY TECHNOLOGY

Key Technologies

By constructing large-scale hierarchical data sets, Tetras.AI's scene classification technology can effectively improve model generalization ability and quickly build training sets for different label requirements. We provide hierarchical model deployment solutions for different hardware platforms.

Scene Understanding

Conduct detection and semantic analysis of photos/videos to generate scene information
Behavior Understanding

After importing the image/video, the label information is generated based on the content understanding of the image
Portrait Verification

After merging into the image/video, portrait detection is carried out according to the content of the image, and character information is generated

APPLICATION SCENARIOS

Application Scenarios

Scene Classification product is applicable to various mobile phone platforms, including:

intelligent photo album management and picture classification function
video and picture recommendation and content search function
personalized video editing functions