Today the majority of ranches are small-sized family operations. In fact, 98% of US farms and ranches are family-owned, accounting for 86% of farm production. Given the popularity of micro-farms, it is increasingly common for workers to have second jobs off their farms to sustain their families. OneCup is committed to helping these ranchers by having its BETSY technology watch the cattle while they are away. The company is partnering with SuperAnnotate to build datasets for training the world's best artificial intelligence to be the rancher's eyes and hands.
TLDR; what the team at OneCup loved most about SuperAnnotate’s end-to-end solution for ML pipelines:
- 10-20x reduction in annotation cost: SuperAnnotate was able to reduce OneCup’s data annotation costs by 10-20x.
- Advanced tooling for higher-quality annotations: By leveraging SuperAnnotate’s feature-packed toolkit, OneCup managed to create more accurate datasets for its AI, BETSY.
- 32x faster quality check: Quality assurance sped up by 32x for BETSY’s training datasets.
- Endlessly scalable curation system (Explore): Using superAnnotate’s curation system, OneCup monitored dataset quality and effectively detected annotation biases for thousands of items before they were set for training.
- 5-10x increase in performance: SuperAnnotate made BETSY’s job 5-10x easier due to the increased quality of annotations.
Bringing automation to precision ranching
BETSY (Bovine Expert Tracking and Surveillance), OneCup's artificial intelligence, uses cameras to monitor and track the health, growth, nutrition, safety, and phenotypes of individual animals. BETSY's computer vision-based 360 live ID platform leverages an animal's unique appearance where animals amass (a feeding station, watering hole, or salt lick) to generate permanent IDs. The system then uses these IDs to alert the farmer when an emergency response is required, whether a calving event or disease detection.
Tracking every individual parameter of each species requires different vision-AI models: That makes BETSY essentially a blend of AI systems linked together with one of the most complex pipelines in the world, according to NVIDIA. BETSY's pipeline complexity required massive training sets with hundreds of thousands of dataset items to annotate/classify — and here’s where OneCup's and SuperAnnotate's paths crossed.
Problem
OneCup has one of the biggest video collections in the world for cattle. Annotating animal data requires utmost precision in tooling, an intuitive interface for faster annotator performance, and a solid curation system to visualize, filter, and quality-check high-volume datasets (not to mention security maintenance across the way). Since having all these components in one place is a hard-find, often expensive, companies tend to opt for having a few separate solutions for each stage in the dataset creation process. SuperAnnotate takes that headache away, offering a single, affordable, and all-in-one platform for your model needs.
Finding a solution with SuperAnnotate
"Today, about 7% of cattle that are conceived don’t survive until weaning. In other species, it's as high as 20%. That equates to several million premature fatalities every year in the US. We think we can make a small difference, we'd be saving hundreds of thousands of lives. We're doing it today for about 50+ different species and breeds, using SuperAnnotate," Geoffrey Shmigelsky, the CTO of OneCup, mentions.
BETSY tracks everything the animal does through SuperAnnotate's bounding box, keypoint, and classification tooling. The robustness of the platform allows OneCup to perfectly identify even the slightest detail in the image/video, such as an animal drop tag.
Animals often hide their injuries, so it's hard for ranchers to catch them. With BETSY, they can spot them in a matter of seconds, which is an indicator of outstanding dataset quality and model performance.
Results
Today OneCup takes advantage of SuperAnnotate's all-in-one AI solution to power BETSY's multi-layer pipeline. "We've tried different products out," Shmigelsky recalls, "we really love the features of SuperAnnotate because we can do things here that we can't do elsewhere." SuperAnnotate serves as a foundation for BETSY's vision technology helping them capitalize on the following:
Advanced tooling: OneCup uses SuperAnnotate's classification, keypoints, and bounding boxes to annotate animals. The ability to label data and multiple metadata keys and values in a bounding box enables the team to do multiple classifications on one piece of data, allowing them to define the age, gender, and activity of cattle all in a single bounding box. That, coupled with keypoints and seamless workflow connections, creates a huge gain in BETSY's training ability.
Comprehensive data format: SuperAnnotate's unique data format carries a lot more information when moving from A to B. As OneCup reveals, it has made BETSY's job "massively easier, up to 5-10x easier because the dataset is so rich in detail when you export it. Other systems don't have that: they follow the COCO format, which lacks the fidelity that we need when we wanna migrate between platforms."
Cutting-edge curation system: A good curation platform is a guarantee for healthy datasets, hence, performant AI. OneCup secures BETSY's success in predictions by making sure every item in the dataset undergoes monitoring and quality check in SuperAnnotate's Explore (curation). Taking a step further and leveraging advanced filtering in Explore, the data team at OneCup recorded first-of-its-kind 32x faster quality assurance. That alone gives OneCup a competitive advantage winning off more time to market BETSY as opposed to using that time to manually inspect datasets.
Cost savings: The combination of faster quality assurance, an inclusive data format, and advanced annotation and curation technology allows OneCup to cost-effectively deploy its artificial intelligence in ranches across the globe. "We got basically a 10-20 fold improvement in cost. What used to cost us tens of thousands of dollars to do, we can do today for several thousand."
Key outcomes
SuperAnnotate helped OneCup reduce data annotation costs by 10-20x and speed up the quality assurance process by up to 32 times. The intuitive and user-friendly toolset made it 5-10x times easier for the annotators to complete their tasks. The intelligent curation system enabled OneCup to create more precise datasets for BETSY and increase its model accuracy.
Do you also want to build the highest quality training datasets and scalable AI pipelines? Join 100+ companies to build the next big thing in AI or become a market leader in your space the way OneCup did.
