Manual behavior scoring is highly time-consuming, often requiring hours of work and multiple team members for validation. As analysis extends over long periods, human factors like fatigue, attention span, and stress can significantly impact consistency and accuracy. This not only increases labor costs—sometimes requiring dedicated staff for scoring—but also reduces data reliability. Automating this process enhances efficiency, consistency, and overall research quality.

Our modules are fully no-code, meaning you can use them from start to finish without writing a single line of code. This is crucial because programming skills are often a barrier to efficient and advanced data analysis. By eliminating this requirement, we ensure that high-quality data analysis is accessible to all researchers, regardless of coding experience.

While open-source software is technically free, the time and labor required to implement and extend it can be extremely costly. In particular developing behavior recognition features can take years, which translates to significant investment in wages and resources. TOMB-AI eliminates this burden by providing a ready-to-use, high-performance solution.

TOMB-A leverages advanced AI-driven tracking and behavior recognition to deliver precise, scalable, and user-friendly analysis, effectively overcoming many of the limitations seen in other commercial solutions. Unlike systems that rely on pixel subtraction for subject identification, TOMB-AI is significantly less affected by environmental variables such as lighting conditions, camera movement, or setup changes. Its AI capabilities enable highly accurate identification and tracking of multiple unmarked subjects within the same environment — a feat that most commercial alternatives struggle to achieve. Furthermore, while many competing systems fail to genuinely recognize behaviors and instead rely on proxies such as region entries, key point confirmations, or subject outline crossings, TOMB-A provides true behavior recognition, reducing the risk of false detections and improving the reliability of analysis.

The number of videos we require varies somewhat dependent on the complexity of the behavior but generally we only require 5 to 7 recordings, each lasting 2-10 minutes of different subjects. Within these videos we need to see the behavior of interest occur. If we are building a model to recognize multiple behaviors at once then we need a good balance of each of the behaviors occurring.

At Metofico we constantly update our software based on customer feedback and market demand, keeping our solutions up-to-date with advancements in the field.

Our team takes care of all the training—simply share videos of the behaviors you're interested in, along with their corresponding scores, and we'll handle the rest.

All analysis data can be downloaded as an Excel file. Each individual metric is also available for download as a PNG or SVG, enabling customization of colors and styles.