Sign in Self service platform for acceptance

Data Annotations

It is suitable for large-scale image, video, voice, text and other special data data cleaning, evaluation, extraction and special information annotation. The professional annotation team provides data annotation services efficiently and stably.

Application service

Data crowdsourcing service

Application service
Product Overview Use Scenarios Product advantages Cooperation cases Related recommendations Customized service

Product Overview

We provide customers with professional AI data annotation services. With Baidu Intelligent Cloud's 10 years of data service experience and rich annotation manpower in the annotation base, we can accurately, efficiently and safely complete various types of data annotation tasks and help customers train algorithm models.

Use Scenarios
  • computer vision
  • Language recognition
  • Natural semantics

  • Image semantic segmentation

  • Picture classification

  • Picture box selection

  • Facial bone management

  • 3D point cloud

  • 2D3D blend annotation

  • Continuous frame annotation

  • Video classification

  • Video content extraction

Image semantic segmentation

Image semantic segmentation is based on polygon annotation of regions, which divides complex and irregular images into regions and marks corresponding attributes to help image recognition model training. It is mostly applied to human body segmentation, scene segmentation and automatic driving road segmentation, and can be applied to intelligent driving, intelligent equipment, and intelligent security scene landing.

10W Area/day

Dimensioning ability

98% +

Accuracy

Application service

Picture classification

Based on the Baidu label base, the human resources can realize the cleaning and classification of tens of millions of images. According to your needs, you can classify the image sets you provide, help image recognition model training, and can be applied to smart retail, smart devices, smart entertainment and other scenes.

300W Figure/day

Dimensioning ability

99% +

Accuracy

Application service

Picture box selection

Picture framing can help image recognition model training, and is used to frame the recognition subject targets in the pictures. It is commonly used to frame faces, human bodies, obstacles, traffic lights, and can be applied to the landing of intelligent driving, intelligent security, and intelligent devices

10W Box/day

Dimensioning ability

99% +

Accuracy

Application service

Facial bone management

Facial skeleton dotting is a point based annotation, which is mostly used to annotate facial features, human skeleton key points and car tire grounding points in pictures, to help image recognition model training, and can be applied to intelligent driving, intelligent equipment, and intelligent security scene landing.

15W Figure/day

Dimensioning ability

98% +

Accuracy

Application service

3D point cloud

3D point cloud annotation can help the training of automatic driving models. Baidu, based on its rich experience in automatic driving annotation and advanced annotation tools, can frame 3D obstacles and semantic segmentation of radar maps, help vehicles better perceive the road surface, and can be applied to the training landing of automatic driving scenes

40W Box/day

Box selection ability

eight hundred Frame/day

Segmentation capability

Application service

2D3D blend annotation

2D3D fusion annotation can help the training of automatic driving models. Baidu, based on its rich experience in automatic driving annotation and advanced annotation tools, can annotate the data of 2D3D multi-sensor fusion at the same time to help vehicles achieve visual and radar perception, which can be applied to the training landing of automatic driving scenes

10W Box/day

Dimensioning ability

98% +

Accuracy

Application service

Continuous frame annotation

Continuous frame annotation is often used for training automatic driving and video image recognition models. By extracting frames from the video and marking the target objects in each frame, it can be applied to intelligent driving, intelligent security, and landing of intelligent devices

25W Box/day

Dimensioning ability

98% +

Accuracy

Application service

Video classification

Video classification is to classify videos by theme by watching video clips to help build a video database. It is commonly used for image recognition model training in the video industry and can be applied to the landing of smart entertainment scenes

1W Segment/day

Dimensioning ability

98% +

Accuracy

Application service

Video content extraction

Video content extraction is to extract frames from videos, transcribe subtitles in each frame, summarize and extract video themes, help build video databases, and is commonly used in image recognition model training in the video industry, which can be applied to the landing of smart entertainment scenes

5W Item/day

Dimensioning ability

98% +

Accuracy

Application service

  • Voice cleaning

  • phonetic transcription

  • Speech segmentation

  • Phoneme annotation

Voice cleaning

Voice cleaning uses technology to clean empty audio, which is monitored manually to screen out qualified audio. Based on Baidu label base, manpower can realize massive audio cleaning, help voice recognition model training, and can be applied to smart home, smart devices, smart customer service, smart stores and other scenarios

three hundred Hour/day

Dimensioning ability

98% +

Accuracy

Application service

phonetic transcription

Speech transcription is based on the content of audio playback and transcribed into the corresponding text. It is commonly used for speech recognition model training. It can support speech transcription in Mandarin, dialect, English and small languages, and is applied to the landing of smart homes, smart devices, smart customer service, smart stores and other scenarios

fifty Hour/day

Dimensioning ability

98% +

Accuracy

Application service

Speech segmentation

Speech segmentation is used to monitor long audio, mark the starting point of speakers in audio, train speech recognition models, and apply to the landing of smart homes, smart devices, smart customer service, smart stores and other scenarios

two hundred Hour/day

Dimensioning ability

98% +

Accuracy

Application service

Phoneme annotation

Phoneme annotation is used to monitor audio, transcribe text and annotate phonetic symbols of text, which is commonly used in speech synthesis technology

five thousand Sentence/day

Dimensioning ability

98% +

Accuracy

Application service

  • Text cleaning

  • Text classification

  • Text enrichment

  • OCR transfer

  • Emotional annotation

  • NLP callout

Text cleaning

Text cleaning is to filter the text according to your rules and select the data that meets the requirements. Based on the manpower of Baidu label base, it can achieve the cleaning of tens of millions of texts, help NLP model training, and can be applied to intelligent customer service, intelligent finance, intelligent driving and other scenarios.

100W Item/day

Dimensioning ability

98% +

Accuracy

Application service

Text classification

Text classification is the attribute classification of text according to your rules. Based on Baidu label base, human resources can realize the classification operation of millions of texts, help NLP model training, and can be applied to intelligent customer service, intelligent finance, intelligent driving and other scenarios.

20W Item/day

Dimensioning ability

98% +

Accuracy

Application service

Text enrichment

Text enrichment is to write text around the theme, so that for the same theme, text expressions are diverse and practical, which can help NLP model training, and can be applied to intelligent customer service, intelligent finance, intelligent driving and other scenarios.

2W Item/day

Dimensioning ability

98% +

Accuracy

Application service

OCR transfer

OCR transcription is to mark and transcribe the text content in the picture. It supports image transcription in Chinese, English and small languages, helps image and text recognition models, and can be applied to smart entertainment, smart devices and other scenes

20W Item/day

Dimensioning ability

98% +

Accuracy

Application service

Emotional annotation

Emotional tagging is to judge the emotional tendency of text expression, classify positive and negative texts, help NLP model training, and can be applied to smart home, smart entertainment, smart finance and other scenarios

10W Item/day

Dimensioning ability

98% +

Accuracy

Application service

NLP callout

NLP annotation is the annotation of text syntax, including slot extraction, text relations, etc., which can help NLP model training and can be applied to smart home, smart entertainment, smart finance and other scenarios

5W Item/day

Dimensioning ability

96% +

Accuracy

Application service

Product advantages

Experienced

10 years of industry and Baidu internal project experience
Efficient completion of various labeling tasks

Leading technology

The strongest labeling algorithm capability in the industry, fully improved
Quality and capacity

Adequate resources

The largest data annotation base in the industry, with sufficient
Mark manpower

Strict standards

Establish the most rigorous safety/audit/operation flow in the industry
To ensure safety and quality

Cooperation cases

Classification and annotation of flower pictures

Project requirements

The 3300000 photos of flowers that have been preliminarily classified automatically are further classified manually.

Annotate results

In 22 days, 3380796 pictures were classified and labeled, with an accuracy rate of 96%.

Marking of face photos

Project requirements

Classification, detection and positioning of 3 million face photos, some of which need to be marked with more than 600 positioning points.

Annotate results

In 27 days, 3 million face photos were marked, with an accuracy rate of more than 99% and a cost of 1/4 of that of outsourcing companies.

Road picture traffic element box selection

Project requirements

500000+road pictures for content entity annotation. The marked entities are various types of traffic elements, including: cars, buses, trucks, vans, pedestrians, bicycles, tricycles, motorcycles, trolleys and other ground elements, and attribute marks are made for elements that are blocked or truncated in the picture. In addition, the traffic lights in the pictures need to be marked separately and distinguish the traffic light attributes (shape, color, direction, etc.).

Annotate results

Delivered in batches, the accuracy rate is 99%, and the partner affirms the delivery speed and quality.

Autopilot dataset annotation

Project requirements

3D point cloud data annotation, some single frame images contain up to 162 vehicles or 80 pedestrians; At the same time, the outdoor dense point cloud data corresponding to the road section should be accurately segmented and labeled according to 19 classifications.

Annotate results

Within 20 days, 22344 frames of image semantic annotation and 1.5km point cloud segmentation results with an accuracy rate of more than 98% were delivered. The work efficiency was twice that of the demander's own annotation personnel, and seven annotation rule changes were handled with high flexibility and professional quick response ability.

Voice data transfer and cleaning

Project requirements

Text escape of 10000 voice data and classified cleaning of voice recording quality (clear, noisy, incomplete voice, etc.)

Annotate results

In 22 days, 3380796 pictures were classified and labeled, with an accuracy rate of 96%. Mark 10000 voice data within 3 days, and the pass rate is 100%

Related recommendations

data acquisition

Secure and efficient data service for solving model algorithm

details

Integrated intelligent driving data solution

Provide a complete set of intelligent driving data services to create a data driven closed loop

details
Customized service

Professional AI data helps the development of enterprise intelligence

Application service
About Baidu crowdsourcing
Application service
Self-Help Platform
Management Console
Friendly Links
Baidu public beta
Baidu MTC
Baidu Brain
Baidu AI market
Baidu Smart Cloud
Dazhou Data Trading Platform
Zhangqiu Data Trading Platform
Linfen Data Trading Platform
contact us

Email: zhongbao@baidu.com

Agency cooperation

Email: Crowdtest@baidu.com

WeChat official account

© Baidu   Jing ICP Certificate No. 030173 Read Before Using Baidu    Service Agreement    Privacy Policy