Drones and computer vision for utility inspection
Sharper Shape uses computer vision models in advanced aerial sensor systems to power the automatic collection and analysis of unmanned aerial inspection data. They built a robust ML pipeline with Labelbox and Valohai.
How Sharper Shape built a robust ML pipeline with Labelbox and Valohai
Sharper Shape creates technology for safe, efficient transmission and distribution solutions for utilities by using drones to perform utility inspections. The company uses computer vision models in advanced aerial sensor systems to power the automatic collection and analysis of unmanned aerial inspection data. Common use cases for their technology include the identification of dangerous setups with electric wiring, such as vegetation growing too close, broken insulators, and more, so that utility companies can find and address potential hazards.
As a company fueled by AI, Sharper Shape sets itself apart with a strong, established pipeline for developing their ML models. While many ML teams rely on glued-together systems, preliminary datasets, and random experimentation, Sharper Shape saves time and engineering costs through a training data platform, Labelbox, complemented by an end-to-end MLOps platform, Valohai. By using these platforms that enable one cohesive process, the team can collaborate more efficiently and take advantage of the many features offered by both platforms to manage a mature and large-scale machine learning operation.
Distribution line components with chipped porcelain insulators and spliced conductors shown with Labelbox UI.
Training multiple computer vision models requires a vast amount of accurately labeled images. Sharper Shape turned to Labelbox to streamline the labeling process, enabling them to use an array of data types, including tiled imagery, and organize their existing data.
Prior to Labelbox, the Sharper Shape team relied on heavily manual workflows and experimented with open-source labeling tools which did not provide the required amount of configuration needed for their needs. With Labelbox, the team could now connect their raw data programmatically into Labelbox via a simple API. Labelbox’s collaboration features also enabled rapid onboarding, training, and throughput for both internal and highly skilled external labelers to work together in one centralized environment. As an upcoming initiative, the Sharper Shape team will be looking to accelerate their labeling process even more with model-assisted labeling which allows teams to import their model into Labelbox and address edge cases.
With the streamlined design of Labelbox, we are able to cut costs on labeling by as much as 50% while maintaining the highest quality in our training data, and get to training our models faster. With human-in-the-loop model-assisted labeling, we expect another huge reduction in time and costs to the labeling process. After a preliminary model is trained, we can run a loop to generate labels from our model’s inference, and feed those back into Labelbox, effectively cutting the labeling load of our labelers to that of reviewing for false positives. That allows us to increase our capabilities and model accuracies exponentially with respect to time for the amount of components and defects we can detect and classify.Edward Kim – Data Analyst / AI, Sharper Shape
After their data is fully annotated inside of Labelbox, the data will be exported to the Valohai MLOps platform where the Sharper Shape team runs their machine learning experiments and training pipelines.
Valohai enables Sharper Shape to train their models on powerful cloud hardware without DevOps support and to house all their collaborative experiments under a single application. Established ML processes can be fully automated into Valohai pipelines so models can be trained each time when new annotated data is available from Labelbox.
Previously, each data scientist would spend up to a third of their time on infrastructure and experiment management, but with Valohai that is all automated. Additionally, as the team has grown from just a few data scientists to a team of four, each new member can be onboarded in a quarter of the time as all the experiments are shared and there isn’t a learning curve to utilizing hardware resources.
Visualization of the MLOps pipeline
Before using the Valohai and Labelbox platforms, we struggled with managing our training data creation infrastructure and manual experiment tracking. We’re now able to concentrate on model building and deployment, without sparing engineering effort and are able to speed up model training by over 10X.Jaro Uljanovs – Data Science / AI, Sharper Shape
Schema of model-assisted labeling
We’re inspired to help companies like Sharper Shape enable their breakthroughs and to make energy grid inspections safe, fast, and affordable.
To hear more about how Sharper Shape utilizes Valohai, reach out to us.
To learn more about how Sharper Shape creates life-saving ML models used in the energy and utility industry, visit www.sharpershape.com
The best of the best
MLOps in the WildA collection of MLOps case studies
The MLOps space is still in its infancy and how solutions are applied varies case by case. We felt that we could help by providing examples of how companies are working with tooling to propel their machine learning capabilities.
Think of this as a lookbook for machine learning systems. You might find something that clicks and opens up exciting new avenues to organize your work – or even build entirely new types of products.Download