Wenn & 3LC’s Pioneering Partnership

At 3LC, we’re ecstatic about the success stories that come our way, showcasing how our products and services enhance and streamline businesses. One of our most impactful collaborations has been with Wenn ASA, a leader in AI technology for the car rental industry.

A New Era in Car Rentals

In the bustling hallways of Munich Airport, something transformative is happening. The traditional process of car rentals, often mired in inefficiencies, misunderstandings, and manual checks, is undergoing a radical change spearheaded by Wenn ASA, a leading AI tech firm dedicated to refining the car rental experience. Their state-of-the-art AI damage detection model provides unparalleled efficiency and transparency, revolutionizing the process from start to finish.

However, as part of Wenn’s powerful AI lies another innovation — 3LC. A groundbreaking product that has seamlessly integrated with Wenn’s operations, enabling it to achieve incredible results.

IOU: Label bounding box vs model inference.

The Challenge: Accuracy and Efficiency

Manual analysis and training improved the AI damage detection model. However, this is too costly and time-consuming. Needing to speed up this process, Wenn partnered with 3LC. The aim was to increase the prediction quality with improved cost and time efficiency. The labeled data available was exact, but more nuanced training data was required to recognize the damage more accurately. Enter 3LC.

3LC: The Game-Changer

3LC offers a revolutionary Python library and service, enhancing model training workflows through data-centric iterative processes. It integrates seamlessly with mainstream ML frameworks, recording per-sample per-epoch metrics, allowing for real-time, interactive visual diagnosis and data editing. This in-depth, granular insight is what Wenn needed.

How Did 3LC Assist Wenn?

1. 1. Fake Damage Classification: By examining the false positives, it was evident that the model was erroneously predicting damages that didn’t exist, often with high confidence. By introducing a “Fake Damage” class, the model could effectively distinguish between genuine damages and potential false positives.
   2. Diversifying Damage Labels: We diversified labels into 16 different damage classes. This categorization allowed the model to understand variations more precisely, and by weighing the classes based on distribution, 3LC’s data loader would guarantee a balanced data stream.
   3. Iterative Training: While retraining the model, we used 3LC’s insightful interface to incorporate data modifications.

The results were an improved model that continues to improve rapidly.

The outcomes were notable:

• • Reduction in false positives.
  • Boost in True Positive accuracy, enhancing the trust and reliability in the AI damage detection process.

Embedding Space: Selection of a cluster of images identifying wheel damages

An Enhanced Customer Journey at Munich Airport

Wenn’s success is more than just theoretical. The implementation at the Munich Airport exemplifies how AI can redefine user experiences. A state-of-the-art automatic damage scanning tunnel, underpinned by Wenn’s enhanced AI, guarantees objective and transparent damage assessments for rental vehicles.

Why This Matters for the Broader Industry

The transformation at Wenn demonstrates the power of 3LC and the potential of data-centric workflows in real-world scenarios. When AI models are trained effectively, they don’t just improve accuracy; they revolutionize businesses, enabling them to scale and adapt in a rapidly changing environment.

Harness the Power of 3LC

At 3LC, we believe in the potential of data. Whether you’re working with Computer Vision models or fine-tuning Large Language Models, our platform offers unparalleled control, transforming ML processes into open, interactive experiences.

Ready to reimagine your model training? Discover the 3LC difference.

Filtering: Many ways to visualize, slice and reduce data points to find problematic data.