Trusted by cutting-edge AI development teams working towards AI-ready implementation.
Datasets
Three collections,
built to train what's next
01 / ROBOT
TRAINING DATA FOR
Robot Foundation Models
REC 01 — ARM CELL
Manipulation Data
10k h
+
Teleoperated pick-and-place, assembly, and tool-use episodes captured with robotic arms, with multi-view RGB, depth, joint angles, and gripper states precisely time-synchronized at the frame level.
Factory
Kitchen
Office
REC 02 — HEAD MOUNT
Ego-centric Data
5k h
+
First-person, head-mounted recordings of real-world tasks — capturing hand-object interactions, gaze, and task narration to train models to understand procedures from the worker’s perspective.
Laundry
Kitchen
Bathroom
REC 03 — HANDHELD
Human Demonstration Data with UMI
5k h
+
Hand-held UMI gripper demonstrations that extend data collection beyond the robot cell, capturing the same action space as the deployed policy through human demonstrations at the actual worksite.
Inquiry required
02 / LANGUAGE & SPEECH
TRAINING DATA FOR
LLM / VLM / SFM
SERVICE
Audio Data Collection
Speech data collection tailored to your AI development specifications — from cast selection, recording equipment, and environment setup to on-site capture and studio-grade recording sessions. Through our data partners’ platforms, we can also source speech data across a wide range of genres.
USE CASE
Developing anomaly-detection systems from equipment sound
Training data to detect equipment failure and abnormal conditions early. From microphone placement and measurement conditions through to recording method, we design the collection to match the site.
Provided data
Paired datasets of normal-operation and abnormal-condition machine sound and facility noise, recorded on site under controlled security conditions — including rarely occurring abnormal sound that is hard to capture in the wild.
EXAMPLES OF AUDIO DATA WE CAN COLLECT AND PROVIDE
Japanese media speech content
Multilingual & parallel speech corpus
Speech dataset with emotional expression
Professionally recorded speech dataset
Ambient & environmental noise dataset
Japanese conversational speech dataset
SERVICE
Text Data Collection
6 DATASET TYPES
SPECIALIST-AUTHORED
RIGHTS-CLEARED LONG-FORM
Text data creation and collection tailored to your specifications. FastLabel assigns domain specialists in finance, medicine, and other fields to create data requiring deep subject-matter expertise, while our media partners provide access to rights-cleared long-form text, including paid articles.
01
Long-form text dataset for finance & economics
FINANCE
04
Learning-material dataset graded by level
EDUCATION
02
Long-form text dataset for medicine
MEDICAL
05
Long-form text dataset for lifestyle & gourmet
LIFESTYLE
03
QA dataset on programming languages
CODE / QA
06
Multi-turn expert conversation dataset
DIALOGUE
03 / WORLD MODELS
TRAINING DATA FOR
World Models
SERVICE
Image Data Collection
Diverse image data sourced from trusted, rights-cleared sources — including stock photo services, specialist media, and trade publications — with captioning and data preparation to make every image training-ready.
USE CASE — FACE RECOGNITION
※ SAMPLE IMAGES ARE FOR ILLUSTRATION
#1
Improving the accuracy of face-recognition systems
Training data designed to improve real-world recognition accuracy for AI face recognition systems — robust across a wide range of scenes, environments, and conditions.
#2
Provided data
High-quality face image datasets covering diverse variations in age, gender, ethnicity, facial expression, lighting, pose, and viewing angle.
EXAMPLES OF IMAGE DATA WE CAN COLLECT AND PROVIDE
Japanese & international face image dataset
Japanese nature & landscape image dataset
Medical scan image dataset (X-ray / CT)
Business form & document image dataset
Consumer-shot furniture & interior image dataset
Specialist media image content
SERVICE
Video Data Collection
Video data planned and captured to your AI development requirements — from cast selection, camera and framing setup to location scouting. We also provide frame-level annotation and labeling, delivering training-ready video data.
CAM 01 — TRACK & FIELD / 60fps
#1
Developing sports motion-recognition and analysis systems
Training data for accurate motion analysis and performance evaluation — covering diverse sports, athlete movements, and environmental conditions.
CAM 02 — SWIMMING / HIGH-SPEED
#2
Provided data
Video datasets covering track and field and swimming events, with diverse camera angles, athletes, and movement patterns.
EXAMPLES OF VIDEO DATA WE CAN COLLECT AND PROVIDE
SPORTS
Sports motion video dataset
PEOPLE FLOW
Video dataset for people-flow analysis
DRIVING
Driving dataset from in-vehicle cameras & LiDAR
