Egocentric Multi-Camera Stereo Dataset (Front + Rear) for Robotics & Physical AI
A multi-view egocentric dataset captured using a 4-camera stereo rig (front and rear-facing), designed for training advanced robotics and physical AI systems. Enables robust depth perception, spatial awareness, and full-scene understanding from a first-person perspective.
Advanced Robotics Dataset – POV, Depth & Thermal (1000 Hours, Egocentric)
1000 hours of real-world egocentric RGB, depth, and thermal data built for robotics and physical AI. Perfect for training multi-modal perception, navigation, and interaction models in complex environments. Designed to accelerate development of reliable, real-world-ready autonomous systems.
Natural Japanese Conversational Dialogue Dataset (High-Quality Casual Speech, Native-Level, Real-Life Scenarios)
A dataset of natural, casual Japanese conversations reflecting real-life interactions, designed to help AI understand and generate fluent, native-level dialogue.
Natural English Conversational Dataset US accent (High-Quality Casual Speech, Native-Level, Real-Life Scenarios)
A dataset of natural, casual American English conversations reflecting real-life interactions in the US, designed to help AI understand and generate fluent, native-level dialogue.
Egocentric POV human-motion video in industrial environments
High-quality egocentric (POV) video dataset capturing real-world human motion in industrial environments. Ideal for robotics, manipulation learning, and physical AI. Scalable production with diverse tasks, rich metadata.
Data for robotics. Embodied AI & Robotics Human Motion Video Dataset. Real-World Human Activity Video. Human Action & Manipulation
Human task execution and motion video dataset for robotics and embodied AI training. Real-world recordings of people performing mechanical tasks and object interactions with detailed annotations for robot learning and human-in-the-loop systems
Multimodal Conversational Dataset for LLM Training (7T Tokens, 200B Messages, 10+ Years, Text, Image, Video, Audio)
Сonversational dataset with text, images, videos, voice messages, and documents. Designed for large language model training, multimodal AI systems, evaluation, and research. Covers 10+ years of public data across 60+ languages with privacy safeguards.
African English Accent Conversational Dataset — Gender, Age, City Metadata with Validated Speech Samples
Validated English speech dataset from six African countries, featuring gender, age, city, and structured audio metadata. Ideal for AI, ASR training, and speech analysis.
+1 more
Latin American English Accent Speech Dataset — Authentic Local Speaker Conversations
High-quality English speech dataset with authentic accents from Mexico, Colombia, Dominican Republic, Costa Rica, Guatemala, and El Salvador. Perfect for AI training, accent recognition, and speech research.
+1 more
Call Center Audio Recordings (100,000+ Hours, High-Quality) in Multiple Languages | Available now (off-the-shelf)
At FileMarket AI Data Labs, we offer an extensive collection of 100,000+ hours of audio recordings from customer-agent conversations in a wide variety of languages. Our dataset meets the highest standards of quality, content, and diversity.