AI Data

Recording Kitchen Work for AI: What Robotics Labs Want From Cooking Video

By Chisom Anen20 November 20267 min read
Recording Kitchen Work for AI: What Robotics Labs Want From Cooking Video

Overview & Key Takeaways

This article explains what AI and robotics researchers look for in cooking video data, focusing on the need for diverse, first-person perspectives from real kitchens. It details how recording your everyday cooking can contribute to advanced AI and what makes this data valuable.

Key Takeaways
  • Robotics labs need first-person (egocentric) video of cooking to train AI systems to understand and perform kitchen tasks.
  • Diversity in cooking styles, ingredients, and environments is crucial for robust AI models.
  • Contributors are paid for approved recordings, with data quality and adherence to guidelines being paramount.
  • TrustAm ensures data authenticity through rigorous verification of contributors, workplaces, and recordings.
  • Nigerian kitchens offer unique culinary practices that are vital for developing inclusive AI.

Cooking is a fundamental human activity, rich with complex actions, ingredient interactions, and environmental awareness. For artificial intelligence (AI) and robotics, understanding these intricate processes is a significant challenge and a key area of research. AI labs developing robots that can assist in the kitchen, or AI systems that can analyze food preparation, require vast amounts of realistic video data. This is where everyday cooks, from home kitchens to bustling buka eateries, can play a vital role.

Why Do AI Labs Need Video of Kitchen Work?

Robotics and AI research labs are actively seeking to build intelligent systems capable of understanding and performing tasks in the real world. The kitchen, with its inherent complexity of object manipulation, sequential actions, and ingredient recognition, serves as an ideal testbed for these advancements. Researchers need to train AI models to recognize ingredients, understand cooking techniques like chopping, stirring, and frying, and even predict the next steps in a recipe. This training data must be as close to real-world conditions as possible to ensure the AI can generalize beyond a controlled laboratory setting.

First-person, or egocentric, video is particularly valuable because it mimics how a human (or a robot with a camera) would perceive the kitchen. This perspective captures the natural hand movements, the view of tools and ingredients as they are used, and the immediate environment. Unlike third-person views, egocentric video provides direct insight into the 'how' of cooking. Projects like Ego4D and EPIC-KITCHENS have highlighted the need for diverse egocentric video, but much of the existing data lacks representation from African kitchens. This is where TrustAm steps in, aiming to fill this gap with authentic Nigerian cooking data.

Nigerian woman wearing head strap with phone camera at a bustling buka kitchen preparing food
Recording authentic kitchen work provides valuable data for AI development.

AI models trained on diverse datasets are more robust and less prone to bias. For example, an AI trained only on Western cooking might struggle to understand the techniques and ingredients used in Nigerian cuisine, such as the preparation of egusi soup or jollof rice. By collecting video from Nigerian kitchens, TrustAm is enabling the development of AI that is more globally relevant and inclusive. This aligns with Nigeria's own ambitions in AI development, as outlined in initiatives like the National Artificial Intelligence Strategy.

What Specific Data Do Robotics Labs Look For?

Robotics labs are interested in detailed, high-quality video recordings that capture the nuances of kitchen activities. They seek data that can be used for a variety of AI tasks, including action recognition, object detection, and human-robot interaction. The more granular the data, the more insights AI models can extract.

Key Elements of Valuable Kitchen Video Data:

  • First-Person Perspective: Recordings made with a head-mounted camera provide the most natural view, showing exactly what the cook sees. This is crucial for training robots to navigate and interact within the kitchen space. The choice of camera placement can significantly impact the data's utility.
  • Clear Action Sequences: Videos should clearly show distinct cooking actions, such as chopping vegetables, mixing ingredients, operating appliances, and plating food. The cleaner the action, the easier it is for AI to learn.
  • Ingredient and Tool Recognition: The AI needs to learn to identify common kitchen ingredients (e.g., onions, tomatoes, spices) and tools (e.g., knives, spoons, pots). Videos that show these items being handled and used are highly valuable.
  • Environmental Context: The background of the recording provides context. This includes the type of kitchen (home, professional, outdoor), the lighting conditions, and the presence of other people or objects. This helps the AI understand that cooking can happen in many different settings.
  • Audio Information: While video is primary, accompanying audio can provide additional cues, such as the sound of chopping, sizzling, or spoken instructions.
  • Language Diversity: Including speech in various Nigerian languages, such as Yoruba, Igbo, or Hausa, alongside English and Pidgin, makes the data more versatile and supports the development of multilingual AI assistants. Sourcing Nigerian language speech data is a key part of TrustAm's offering.
Close up of Nigerian hands chopping fresh tomatoes on a wooden cutting board
Detailed hand movements and ingredient interactions are vital for AI learning.

The goal is to create comprehensive datasets that capture the breadth of cooking activities. Datasets like EPIC-KITCHENS provide a foundation, but they often lack the specific cultural context and diversity found in Nigerian households and food businesses. TrustAm's focus on collecting data from a variety of Nigerian trades, including cooking, helps to address this gap, creating more representative training materials for AI and robotics.

How TrustAm Ensures Data Quality and Authenticity

For AI and robotics teams, the trustworthiness of the data is paramount. They need to be confident that the video they receive is genuine, collected ethically, and accurately represents real-world activities. TrustAm has implemented a robust system to ensure the highest standards of data integrity and contributor privacy.

Our Verification Process Includes:

  • Contributor Verification: Every contributor undergoes identity verification using their BVN or NIN, matched with a live selfie. Payments are made directly to bank accounts or TrustAm accounts held in the contributor's name, ensuring one person, one account.
  • Workplace Registration: Our field agents physically register workplaces, obtaining consent from the owner before any recordings begin. Each location is mapped with a geofence.
  • Recording Integrity: Each recording starts with a unique random code read aloud, with hands shown for context. Motion sensors (IMU) and GPS data are recorded alongside the video to verify movement and location. Automated checks detect issues like frozen video, duplicates, or stationary phones, complementing manual review of each approved minute. This detailed process, which includes checks against fraud, is vital for producing reliable data. Read more about our fraud controls.
  • Ethical Data Handling: Contributors give explicit consent on video in their own language before recording. Bystander faces are blurred, and all participants in conversation tasks consent beforehand. Personal data is stored securely, with options for deletion and future withdrawal of licensing.

This meticulous approach ensures that the data supplied to AI and robotics teams is not only diverse and representative but also ethically sourced and highly reliable. The data is then processed into formats usable by AI labs, such as MP4 video with JSON metadata, and can be delivered in formats like LeRobot or RLDS upon request. This adherence to standards is crucial, as highlighted by the need for comprehensive documentation in datasets, similar to the principles outlined in Datasheets for Datasets.

Recording Your Kitchen Work: What You Need

If you or someone you know is involved in cooking, whether professionally in a restaurant, a roadside buka, or even a passionate home cook, you may be eligible to contribute. The process is straightforward and designed to be accessible.

You will need a smartphone capable of recording video at 1080p resolution and with at least 8GB of free storage. The TrustAm app handles the recording and uploading process. We are currently piloting this in Lagos, focusing on trades like cooking, tailoring, auto mechanics, and more.

By participating, you help advance AI technology and contribute to a more inclusive digital future. Your everyday cooking activities can become a valuable part of datasets used by researchers worldwide, helping to build smarter and more capable robots and AI systems. This contributes to the growing demand for AI training datasets, particularly those that offer real-world diversity.

Frequently Asked Questions

Is recording my cooking for AI safe and private?
Yes, safety and privacy are our top priorities. Your identity is verified, but your personal data is securely stored. Recordings are processed to blur bystander faces, and contributors give explicit consent. You retain control over your data and can request deletion at any time.
What kind of cooking activities are most useful?
AI labs are interested in a wide range of activities, from basic preparation like washing and chopping ingredients to complex cooking processes like frying, baking, and stewing. Any clear, sequential cooking action performed naturally in your usual kitchen environment is valuable.
Will I be paid for my recordings?
Yes, contributors are paid for approved minutes of recording. The pilot program in Lagos pays ₦4,750 per approved hour. Payment is made in naira to your bank account for minutes that pass our quality review process.
What if I cook in a busy market or a commercial kitchen?
Commercial kitchens and busy market stalls are excellent environments for data collection, as they offer unique insights into different cooking practices. As long as the workplace owner consents and the recording process is followed, your contribution is valuable.
Do I need special equipment to record?
A modern smartphone (Android 9 or newer) with a good back camera (1080p recording capability) and sufficient storage is all you need. TrustAm provides guidance on how to use the app and wear the phone securely for optimal recording.
How does this data help AI?
This data helps AI and robotics systems learn to understand human actions, recognize objects, and navigate complex environments like kitchens. It's essential for developing robots that can assist with cooking, food preparation, or even perform these tasks autonomously.
Buy from TrustAm Data

Need AI training data from Nigeria?

Egocentric video of skilled work and Nigerian-language speech, collected to order. Email [email protected] or request a free 10-hour sample; a person replies within one working day.

CChisom Anen

Founder and CEO of TrustAm. Building Nigeria's reputation marketplace for verified service providers. Reviews unlock only after real payments.

Disclosure: This article is published by TrustAm Data, part of TrustAm Limited. Links to /data pages are our own data service.

Sources & References

  1. National Artificial Intelligence Strategy— ncair.nitda.gov.ng
  2. EPIC-KITCHENS— epic-kitchens.github.io
  3. Datasheets for Datasets— arxiv.org
  4. AI training datasets— grandviewresearch.com
  5. Nigeria Data Protection Commission (NDPC)— ndpc.gov.ng

Related Articles