Co-one is proud to have worked with Cybelle Medical supported by Vestel Ventures which is a corporate investment company. Cybelle Medical produces an innovative automatic laser epilation device that uses artificial intelligence to detect hair follicles and shoot directly at the hair root.
The collaboration aimed to support Cybelle Medical in developing their automatic laser epilation device, which utilizes artificial intelligence to detect hair follicles and perform the targeted treatment. The accurate identification of hair follicles is crucial for the device’s effectiveness and user satisfaction. This project sought to overcome the challenges faced in obtaining efficient and precise data labeling services. By providing high-quality data labeling, Co-one aimed to contribute to the development of Cybelle Medical’s innovative product, ultimately improving patient outcomes in the field of laser epilation.
Cybelle Medical faced significant hurdles in acquiring accurate and efficient data labeling services for its automatic laser epilation device. Previous attempts, whether conducted in-house or through outsourcing, failed to achieve the desired level of efficiency and precision. These challenges hindered Cybelle Medical’s progress in developing a reliable and effective solution, underscoring the critical need for a more robust data labeling process.
The data labeling process began with Cybelle Medical sending us images to be labeled. The task involved marking every hair that appeared in the photos, including the point where the hair exits the skin and the hair root. Once the data labeling was completed, we delivered the labeled images to Cybelle Medical and held a meeting to discuss the results.
During the meeting, we received valuable feedback from Cybelle Medical that helped us improve the quality of our data labeling services. For example, Cybelle Medical pointed out that when labeling thin and thick hair data, the distance between the root and the exit point should be evaluated differently. Additionally, they requested that the exit point of thick hairs be marked under the skin, as these hairs are often visible beneath the surface. They also asked not to mark hairs that are too long and out of shape. In addition, they requested that we not mark photos with too many stains on the skin or photos with darkening or light glare due to shooting. These new rules were adopted. We have created a guideline within the framework of these new rules, and we set to work editing the images that Cybelle Medical had sent back to us for re-editing. In total, we labeled over 310,000 hair follicles using these new rules.
The feedback that Cybelle Medical provided was critical to the success of the project. Thanks to their input, we were able to improve our data labeling processes quickly and efficiently. In particular, they saw significant increases in the detection of fine and thick hairs, with a 15% increase in the detection of thick hair and an 18% increase in the detection of fine hair.
We leverage the power of our crowdsources to label data. We get unbiased data labeling by distributing the data to our multiple users. If there is a labeling that cannot be decided by multi-users, it is decided by our experts. The quality of data annotated with cross-validation is under our control.
The partnership between Co-one and Cybelle Medical yielded remarkable improvements in the detection of hair follicles and the overall performance of the automatic laser epilation device. By incorporating the feedback received from Cybelle Medical, Co-one refined their data labeling process to enhance accuracy and efficiency. The project resulted in a significant increase in the detection of both thick and fine hairs, with a 15% improvement in thick hair detection and an 18% improvement in fine hair detection. Cybelle Medical expressed their satisfaction with the results and expressed their intention to continue collaborating with Co-one on future projects, highlighting the value of Co-one’s data labeling services in their ongoing development efforts.
“As a result of the feedback, efficient work was carried out. In total, 310 thousand data labeling was carried out. Increases were seen in the detection of fine and thick hairs. There was a 15% increase in the detection of thick hair and an 18% increase in the detection of fine hair.”