Pixta AI solution
for Multi purpose training model
The customer is an information and communications technology leader in Japan. As the exponential growth of ICT demand, the customer vigorously experiment AI solutions to variety of industries & verticals. They develop a multi training model with a mixture of human & object detection.
To help with input training data for the multi-model, Pixta is required to provide 15,600 Japanese human face images in balance of age, gender; 5,200 images of cars & 5,200 images of airplanes in multiple scenes.
As powered by AI experts, PIXTA AI team deploy computer vision model for data collection curation, sorting & making sure to match with customer’s requirements. Besides, a human-in-a-loop process is applied to shorten delivery time.
26,000 images are delivered in each subset under high satisfaction level. The customer could test & enhance their system to a higher accuracy
The customer is sure that the dataset delivered on time to match with their project schedule, without any concern regarding legality of human data.
To Pixta: Accuracy is maintained at 97% from planning to delivery, involve less human resource to fulfill the task.