Data Labeling Case Study in New Retail — Image Recognition and Labeling
In the Internet age, fragmented consumer behavior makes traditional retail methods unsustainable. New retail as a brand-new retail solution has been widely used.
With the rapid development of AI technologies such as image recognition, new retail companies have generally achieved cost reduction and efficiency enhancement. Goldman Sachs predicts that by 2025, the application of artificial intelligence will save the retail industry an annual cost of 54 billion U.S. dollars and bring 41 billion U.S. dollars in revenue.
More read: What is New Retail?
At present, one of the most popular solutions in the new retail industry is the “visual recognition solution”, which uses image recognition as the technical core, cameras as the hardware core, to detect and classify target products, realize automatic recognition, and improve the shopping experience, while saving labor cost.
This type of solution cannot be separated from the support of a data labeling service. We’d like to share a labeling case of a product display inspection in a new retail scenario to visually show the specific AI application.
1. Annotate objects
Commodities/Price tags/Items out of stock on the front but in stock at the back/No item for sale
2. Labeling type
Image 2D frame
3. Category attributes
Label A (commodity): labeling all categories of goods for sale;
Label B (price tag label): All labels except blank price;
Label C (no item for sale ): Imaging the estimated size of shortage. About empty space, leave the object size to the imagination based on price tag location.
Label D (out of stock on the front shelf but in stock at the back): the front row of goods has been sold while there are goods for sale in the back. Labeling the empty space of the goods for sale in the front.
4. Labeling rules and standards for various objects
a. About stacked products, each one should be labeled (such as instant noodles);
b. About flat products, only the top one needs to be labeled (such as chocolate);
c. Judgment basis: whether the leaked part can determine what kind of product it is;
d. For products positioned inside and outside, only the outermost one should be labeled;
e. The covered object needs to be outlined its full size by imagination;
Outsource your data labeling tasks to ByteBridge, you can get the high-quality ML training datasets cheaper and faster!
- Free Trial Without Credit Card: you can get your sample result in a fast turnaround, check the output, and give feedback directly to our project manager.
- 100% Human Validated
- Transparent & Standard Pricing: clear pricing is available(labor cost included)
Why not have a try?