(MENAFN Editorial) Over the past several years, with the development of big data, Internet, and computer science, the artificial intelligence has entered the picture. Companies in this field have sprung up. As a branch of artificial intelligence, image processing has become a popular choice for some start-ups… HuffPost Interview Link:
Over the past several years, with the development of big data, Internet, and computer science, the artificial intelligence has entered the picture. Companies in this field have sprung up. As a branch of artificial intelligence, image processing has become a popular choice for some start-ups. But Tuputech has made its own way through content reviewing with image recognition. Recently, we had an interview with this company and we talked about image recognition, artificial intelligence, and the future prospect.
Q: Why did you want to found a company in image recognition?
A: With a large amount of contents displayed through images and videos, we have welcomed an era full of images. On the one hand, though common Internet companies have abundant demands of image processing, they don't always have the technical capability of artificial intelligence. Therefore, their demands for image recognition, understanding, and analysis are steadily on the increase. On the other hand, daily data of images and videos generated by users of those Internet giants are extensive, but these giants are too busy to review these data since their primary service is not content review. In the past, reviewing images and videos heavily relies on human. With the data soaring, reviewing by human cannot meet current demands of content review. So we believe there is a promising future of image recognition.
Q: What role is your image recognition service playing? Who are your target customers?
A: We concentrate on image and video recognition. We can detect various contents including explicit contents, violent contents, junk ads, and so on. In order to meet reviewing demands of detecting explicit images in different countries, we would train our AI model accordingly. So we are able to detect explicit images of all skin colors. Except for common detection of explicit images, we are also capable of detecting child pornography.
Our target customers are mainly live-streaming apps, short video platforms, video sharing platforms, and social media. Except adult websites, the appearance of pornography, especially child porn, is not always acceptable on the Internet. Besides, now video dating apps are extensively used, but it is hard to administrate these apps because of the privacy protection. Yet we are not only providing API but also SDK service to detect explicit contents. SDK can be embedded in apps and monitor the video calls in real-time to detect if there is any violation. In addition, as for those video-dating apps among strangers, our service can also identify whether a user is a teenager or not. In this way, we can prevent teenagers from using such apps as much as possible.
Q: How can you apply your technology in the real commercial world?
A: Actually the commercialization of AI takes step by step. AI technology has some special features. It is easy to develop a model with 70% accuracy, but it is not easy to improve the remaining 30% of accuracy. But we can guarantee both the effect of our models and quality of our service. We will conduct iteration to improve the accuracy. In this way, our whole service can be upgraded and expanded.
Q: How do you customize image recognition solution for different companies? What is the biggest difficulty during the customization？ Can you share with us some of your experience?
A: Normally, we follow these steps to customize a solution for our customers:
First, we analyze the feasibility according to a customer's requests. We will dig for the urgent requests, applying scenario, applying method, and desired effect from our customers. And then considering the cost and the joint point of their requests with our technology, we will analyze if it is possible for us to meet their requests.
Second, we design a customized solution including labels, data source, parameters, indices, and using method.
Third, we arrange the schedule of developing the model. We will provide a demo to our customer for testing, optimize this model according to our customer's requests, and then deliver the final model.
Compared with standard solutions, main difficulty in customizing solutions is longer period and higher cost of development because of big difference in individual requests from our customers. And not all of the requests can be met with current technologies.
Compared with standard solutions, main difficulty in customizing solutions is longer period and higher cost of development because of big difference in individual requests from our customers. And not all of the requests can be met with current technologies.
Our experience in this field:
When we analyze the feasibility and customize solutions, we will integrate our customers' requests. On the basis of satisfying their basic requests, we also try our best to standardize our solutions. In this way, we can improve the efficiency of development. Besides, we will sort out prior requests in advance. After meeting prior requests, we will continue to improve our models.
We will make a simple demo in preliminary stage to evaluate the level of difficulty and find potential problems. Our customers can check our technical capability through this demo as well. We will clarify customers' requests and cooperation process to avoid waste of time and efforts.
Q: What's your opinion about the development of AI in next five years?
A: I believe AI will have a substantial influence on every field. Resources and assets will be rearranged accordingly. The impact of AI is unconscious, but also fast growing and far-reaching. From national status to city's development, from the survival of a company to the job-hunting of an individual, the impact of AI is increasing. The world will witness a revolution under this impact. And individuals will face with new opportunities and challenges. Media Contact
Company Name: FINANCE NEWS CO., LTD.
Contact Person: Terry Aziz
Email:
Phone: 01-267-4706518
Country: United States
Website:
MENAFN0912201700703268ID1096211467