Using computer vision to automate the measurement of physical operations for enterprises using their own IP camera network
Industry
Software
Use Case
Video Analytics
Company size
11-50 employees
Headquarters
New York, USA
Data labelling
Discover our solutions
About
Safari AI is a computer vision company automating the measurement of physical operations for enterprise companies.
Founded in 2018 as a curb space availability monitoring company, Safari AI pivoted towards its current mission in 2022. Now, they help large organizations across industries such as retail, logistics, or real estate replace manual, physical or non-existent data collection of mission-critical processes with end-to-end computer vision solutions.
Safari AI's wide variety of use cases includes pedestrian counts, vehicle classification, staff detection, and more.
The challenge
Safari AI is a company that specializes in providing end-to-end computer vision solutions for large enterprises, automating the measurement of physical operations. The company works with customers across various industries, such as retail, real estate, restaurants, logistics, and more. Their main challenge is to ensure the highest degree of accuracy for their vision AI solutions while delivering them in a timely manner.
To meet the demands of their customers, Safari AI needed a robust ML training data platform to annotate data and train accurate models at scale.
The solution
Safari AI partnered with V7 to expedite ground truth creation and enhance their ML model performance. By utilizing V7's labeling workforce and advanced platform features like custom workflows and model-assisted labeling, Safari AI achieved a significant reduction in annotation costs and delivery time, which was cut in half. Leveraging V7's on-demand workforce, Safari AI also improved their operational efficiency with minimal overhead.
The results
Safari AI has been able to scale their operations significantly with the help of V7's intuitive platform and professional labeling services. This has allowed them to build computer vision solutions for their clients faster and more cost-effectively, while maintaining high levels of quality and accuracy.
By cutting down delivery times in half, Safari AI can take on more projects without having to hire additional staff or increase management overhead. According to Adrian Levitt, Head of Operations, "V7 and our labeling partners play a very important role in the development of our products, helping us deal with the surges and lulls of customer pipelines. Thanks to V7, we have a workforce and platform on-demand that can meet our seasonal and sales-driven needs."
Thanks to V7's model-assisted labeling workflows, Safari AI has been able to annotate data faster, improve the quality of their labeled data through a more comprehensive review process, and lower the cost per project.
Training data needs
Safari AI uses customers' existing cameras or their own equipment to capture images and videos, which they label to build object detection models. These models are then deployed on the client's cameras, providing valuable insights, such as store visitor count or peak parking occupancy times.
To build accurate models, Safari AI deals with large amounts of data, sometimes tens of thousands of images, all of which require accurate labeling. Some models require ongoing data inputs to improve their quality over time.
Safari AI uses V7 to outsource the entire image labeling process to V7’s labeling partners. After importing the data to the platform, providing instructions, and setting up the workflows, V7’s labelers annotate all the images with bounding boxes or polygons. Then, Safari AI's machine learning engineers can use them to train and retrain their models.
Why V7?
Safari AI needed an end-to-end training data platform to annotate data and train models at scale quickly. The company's standard project completion timeline is just one month, making it crucial for them to have a trusted labeling partner and a reliable platform like V7 to deliver quality vision AI solutions promptly.
Since Safari AI's labeling needs depend on the number of clients, business types, and seasonality, maintaining an in-house labeling workforce would strain resources. Additionally, the team faces challenges in annotating their data due to varying weather conditions, camera angles, and shadows. V7 solves these problems by providing on-demand, high-quality labeling services and custom model-assisted workflows.
Safari AI now leverages V7's labeling workforce and model-assisted labeling, and they can label more than 10,000 images per week, achieving higher model accuracy faster.