Featured Resource
Featured Case Study
Ancera speeds up pathogen detection model development, regains months of engineering time
Industry
Healthcare | Software Analytics
Region
North America
About
Ancera, a software analytics company for the poultry industry, focuses on improving live production and processing by monitoring and responding to pathogen risk, such as Salmonella and Coccidia, across the supply chain. Using high-resolution imaging systems and computer vision models, Ancera has helped poultry integrators save millions of dollars and gain real-time visibility into which farms face microbial threats and where interventions are needed to minimize downtime.
Key Results
- Development feedback loop cut from multiple weeks to days, enabling faster model iteration and QA
- Over 7% improvement in model performance from using higher-quality training data
“The biggest benefit of using FiftyOne Enterprise has been the speed of development. What used to take weeks or even months can now be done in days, with fewer people and better results. It’s freed up our team to focus on what they do best while accelerating our computer vision pipeline.”
Challenge
Manual workflows and limited model visibility hindered Ancera’s operational scalability.
Ancera’s platform collects and processes large volumes of poultry data from hatcheries, feed mills, breeder farms, growers, and processing units to detect pathogen risks using computer vision models. As Ancera’s imaging operations scaled, manual data and model workflows were slowing development, and the ML team faced several challenges:
- Slow model analysis and visibility: Ancera’s ML engineers lacked an intuitive way to view, filter, and root-cause suboptimal model performance, often taking weeks of effort across multiple people to scroll through large volumes of data to identify and correct mislabeled data.
- Fragmented collaboration: Non-technical stakeholders (microbiologists) couldn’t easily access or interpret the results of their lab tests, without engineering assistance, slowing iterations and feedback.
As a result, deploying new models was taking over 3-4 months instead of a few weeks.
Solution
Ancera adopted FiftyOne Enterprise for their complex visual data and model analysis layer to accurately classify Eimeria species.
Ancera’s data includes distinct data types such as high-resolution microscopy images, DNA sequencing results, and environmental metadata like flock health and growth rates from a variety of sources. For example, fecal samples from live birds or swabs from machinery in processing plants are translated into images. The variability in sample quality and background noise makes data curation quite complex.
The Ancera team integrated FiftyOne Enterprise to streamline data curation by enabling the team to visually sift through the noise, isolate relevant objects, and maintain high-quality training and test datasets across diverse imaging conditions. They incorporated FiftyOne capabilities to find model performance issues using the ability to analyze false positives/negatives, confusion matrices, and model scores.
Key Results
Faster feedback loops and over 7% improvement in model performance.
Using FiftyOne Enterprise, Ancera accelerated the development and performance of their computer vision model for accurately classifying Eimeria species, a pathogen that causes coccidiosis, impacting animal health, productivity, and overall farm profitability.
- The team gained instant visibility into model predictions and label quality. Using the filtering and patch view capabilities, they were able to review and relabel 20,000 detections across hundreds of high-resolution images in just a few days, a task that previously took multiple weeks.
- This improvement enabled retraining the model with higher-quality data, resulting in more than 7% improvement in performance with a noticeable boost in F1 score, precision, and recall.
- Cross-team efficiency: Ancera’s microbiologists could analyze the results of their lab results independently using a browser link to step through the predictions in FiftyOne, freeing the ML team from manual report generation and speeding up QA cycles.
“FiftyOne has helped us shrink our feedback loop from months to days, allowing us to catch issues we didn’t know existed. It’s a big part of our strategy to speed up model development across teams.”
Ready to get started?
Join the community of developers and scientists who are already using FiftyOne to improve the quality of their data and build better models