Welcome to Computer Vision 101, your guide to using the ErgoPlus Platform's video motion capture technology for ergonomic risk assessments. In this tutorial, we’ll show you how to harness the power of cutting-edge computer vision to streamline your data collection process. This technology acts as your "copilot," helping you quickly and efficiently capture the key movements and postures of your workforce. While it significantly speeds up assessments, you remain in control, ensuring accuracy and fine-tuning the data where necessary.

By combining human expertise with AI-driven insights, you can reduce the time spent on manual data entry and focus on making decisions that improve safety and reduce injury risks. Let’s dive in and explore how to get the most out of this powerful tool to transform your ergonomics program.

The most important part of this is to keep the worker completely in the frame of the video for the duration of the video, trying to get a front or side-facing view of the individual where possible.

The technology has been trained to identify and track the joints of the human body and so as a user, your job is to make sure those joints are there in the videos you are taking. If the worker leaves the frame of the video, the computer vision is trying to find them but can’t. And so you’ll see the skeleton in the overlay video desperately searching for worker, and as it does, it can potentially be recording faulty data. We don’t want that, so make sure the individual stays within the frame of the video.

This is especially important at the beginning of the video. This is when the computer vision is focusing in on the individual’s joints. So, if you have to include multiple people in a video or there going to be inevitable obstructions as you shoot the video, make sure you at least start the video with the individual you are focusing on completely in the frame of the video. From there, you can move to an angle with multiple people or with occlusions and you have a better chance of getting a good result from the computer vision.

Right now the file size limitation for videos is 200MB, or about two minutes. If a single task has a longer cycle time than two minutes, you can always break that task up into multiple tasks in the software in order to capture the entire cycle.

Sometimes loose fitting clothing can hamper the computer’s ability to see those joints. Depending on the clothing or the lab gear or the big coat, the looser it is, the less definition there is to the joints and the harder of a time the computer is going to have locating those joints.

The 3motionAI vision technology supports most standard video formats, including MP4, MOV, M4V, AVI, and WMV file types. This means you can take your videos using most modern phones and devices.

There is no strict minimum video length, but we recommend capturing 1-2 seconds of video before and after the movement of interest. For workplace analysis, we suggest capturing 1-3 cycles or up to 2 minutes for optimal analysis.

Our 3D motion analysis achieves over 90% accuracy compared to lab-grade motion capture systems (the gold standard for human motion tracking), assuming the video meets the necessary quality standards for analysis.

Video processing time is proportional to the duration, resolution, and frame rate of the video. In general, video analysis should take under 3-5 minutes.

Yes, videos can be downloaded post-evaluation in the form of our Overlay Video.

We offer various options to maintain privacy and anonymity. For the Overlay Video, faces are blurred by default, and we can blur the entire background to protect work processes. Don’t hesitate to contact Sales or Customer Support to explore all our options.

While the hands are part of our whole-body model, hand-specific activity evaluation is not yet available. A separate hand activity module is currently being developed, including a hand-specific camera view and detailed hand risk analysis.

Videos can be captured with any standard recording device, such as smartphones, tablets, or external cameras. For best results, we recommend 1080p resolution, at least 30 frames per second, and clear contrast between the human and the background.