- Faster Annotation: Structured campaigns and dataset queries multiplied annotation speed by 3.
- Smarter Experiment Tracking: Centralized datasets, models, and experiments reduced weekly time lost by ~20%.
- Breakthrough Insights: Traceability across the model lifecycle allowed the team to identify improvements faster.
- Team Empowerment: Freed the data science team from pipeline maintenance to focus on innovation.
About Ficha
Ficha is an innovative company in the waste management industry. Using cameras and Computer Vision inside garbage trucks and collection bins, Ficha provides cities and housing associations with precise data on recycling quality. Their goal is to pinpoint sorting issues and deliver targeted information to citizens, helping everyone recycle better. Their technology is designed to make waste management smarter, more efficient, and more sustainable.
Ficha's CV Journey: Scaling with Rapid Growth
Computer Vision was at the heart of Ficha’s product from day one. As the company grew quickly and the number of clients increased, the existing AI workflow—based on scripts and manual data transfers—began to slow down the team.
Annotation was hard to organize and track, and fragmented processes meant valuable information was sometimes lost between datasets, models, and experiments. Even with more and more data available, model performance wasn’t improving at the speed required to keep up with the company’s growth.
One of the key lessons Ficha learned was that improving AI performance isn’t only about more data or training cycles—it’s about creating the right structure. By improving pipelines for annotation and experimentation, data science work became much clearer and faster.
“We suddenly had access to a lot of data, but our models were improving quite slowly. Without a strong process, we couldn’t fully benefit from it. Picsellia gave us the structure we needed to iterate faster and keep innovating.”
The Solution: A Single Platform to Streamline the Workflow
Ficha needed more than a patch—they needed a single platform to support their entire AI lifecycle. With Picsellia, they found an integrated solution that brought structure, speed, and traceability to every stage:
- Data: Visualize and query large image datasets with ease.
- Annotation: Organize campaigns and queries, tripling annotation speed.
- Experimentation: Launch, track, and compare experiments with full traceability.
Most importantly, Picsellia gave them end-to-end visibility: the ability to connect a single prediction back to the exact image and annotation that influenced it. This traceability allowed the team to understand issues quickly and continuously improve their models.
Business Impact of Computer Vision with Picsellia
By adopting Picsellia, Ficha transformed not just the speed, but the way the team works and innovates:
- Radical Speed and Efficiency: With annotation x3 faster and ~20% time saved weekly on tracking and management, iteration cycles are dramatically shorter.
- An Empowered and Focused Team: Freed from repetitive maintenance tasks, senior data scientists can delegate effectively, and junior members can contribute to real innovation.
- A Foundation of Trustworthy AI: Standardized annotation and experiment tracking ensure clarity and reliability. Ficha can now trust their data and confidently deliver accurate insights to their clients.
Moving Forward
With a fast, reliable, and scalable MLOps foundation in place, Ficha’s team is free to focus on what matters most: innovating. No longer slowed down by fragmented workflows, they are pushing the boundaries of waste-tech, confident in their ability to build, deploy, and improve AI solutions efficiently.
