Client

• App used by e-commerce marketplaces and D2C manufacturers for finding a good shoe fit online using just a phone camera
• Reduces number of returns for online orders and improves customer satisfaction
• Trusted by famous brands like Nike, Reebok, Adidas, Puma, and Vans

Challenge

• To train the algorithm, photos needed to be taken with a variety of lighting, backgrounds, skin colors, and foot angles
• Collecting photos from developers’ personal networks was too limited
• Verifying data manually was overly time-consuming

Solution

1. 2472 unique videos collected in less than 3 days
2. Videos were reviewed in Toloka to check quality
3. Videos were split into frames to get more than 150,000 total images (3x the target number)

“Toloka helped us tremendously by collecting and labeling over 50,000 images of different human feet for our algorithm. Using this data, we were able to train our neural network which, in turn, made our app’s 3D scanner 12% more accurate!”

- Roman Kutsev, Crowd Solutions Architect, Neatsy