Improving customer parking experience with
High-Quality License Plate Labeling.

Client:

Quercus Technologies is a globally positioned company in the parking sector with facilities worldwide. It is a world reference in license plate recognition, vehicle detection technology, and parking guidance systems.

Its solution relies on algorithms for the detection of vehicle plates to maximize efficiency in the management and handling of driver parking through cameras at parking entrances and in any parking facilities to maximize the efficiency improving the customers’ parking experience.

Challenge:

Quercus faced a period of stagnation where errors in the ML model couldn't be solved by using the same algorithm. Consequently, the performance of the cameras lacked speed and accuracy, especially in challenging conditions.One of the primary obstacles was the need to enhance the data labeling process, essential for both training and evaluating the camera models. Initially, they attempted to address this by labeling the data in-house, but it yielded unsatisfactory results due to a high likelihood of errors and a significant time investment by the team. In addition to the labeling process, the team had to undergo a reviewing and correction process to achieve the expected quality.These situations led them to recognize the need for a data labeling provider. However, their initial experience with such a provider did not significantly improve the errors present in the labels. Thus, the challenge laid in finding a data labeling provider that could deliver data with the expected quality.

Begoña Vizcaíno
CV Research Engineer
Quercus

"We are very pleased because LinkedAI’s team has consistently proven to be proactive. They have been available to address our requirements at all times.Moreover, when an error has occurred from any labeler, it is always taken into account for the subsequent project executions. Any doubts that arose within the labeling team were consulted with Quercus to avoid making assumptions in the labeling, contributing to preventing future errors."

Quercus recognized LinkedAI as the right partner for
three main reasons:

1) Expertise from the specialized labeling team.

The images provided by Quercus include pre-labels for validation of bounding boxes and the classification of 74,794 images. This process was conducted according to various parameters established by Quercus, which emerged as a result of previous meetings. The parameters are modified on the fly based on the client's needs.Understanding the most significant challenges, evaluating the state of the AI model at the time, and thoroughly grasping the client's needs are the starting points for LinkedAI in its quest to provide the best service.

2) Delivery times for achieving objectives and project development.

Since the start of the service in the year 2022, LinkedAI has consistently delivered the validated datasets within the established deadlines, even when changes to specifications were necessary during the labeling process. This certainty allows Quercus to organize its projects internally without being hindered by the need for labeling data.

3) Quality in labeled data.

LinkedAI has real-time tracking tools for client clarity on project development. However, in Quercus's case, according to Begoña, their use is not necessary due to the trust developed in the LinkedAI team through continuous feedback received from the assigned Project Manager. The smooth communication and willingness to make any changes during the project, or in the output format of the labeled data, have instilled confidence in the quality of LinkedAI.

Impact:

Quercus assessed the results in terms of quality and communication, noting a significant impact, especially in terms of efficiency, peace of mind, and confidence in the direct use of the labels in the model.In addition to the positive aspects of time savings for the Quercus team, the most notable gain is reflected in the confidence gained by receiving labeled data of a higher quality than expected and within the established timelines. This achievement has led to a significant reduction in errors in the model, restricting them to very specific modifications as the project scales.

Begoña Vizcaíno
CV Research Engineer
Quercus

"I am very pleased to work with LinkedAI; communication has been comfortable and easy for me. I emphasize the punctuality in delivery dates; even when requesting changes midway, everything has been delivered within the established timelines. Also, the quality of the data we receive from LinkedAI and their consideration of specifications for future projects. In general terms, it has saved us a lot of time and provided us with a great deal of peace of mind."