Overview
- Chinese giant Baidu has launched its own automated machine learning platform
- Nowadays, EZDL allows to build two types of models: image classification and object detection.
- The construction and implementation of models is a process of 4 steps and takes less than 15 minutes on average.
Introduction
Automated machine learning has carved out a huge space in the industry and shows no signs of slowing down. Led by the Google Cloud AutoML platform, most of the big tech companies have released their own tools to spice up the competition. AutoML can no longer be considered a fad, is extremely useful and is here to stay.
The latest release comes from Chinese giant Baidu, que anteriormente ha sido noticia por avances en el deep learningDeep learning, A subdiscipline of artificial intelligence, relies on artificial neural networks to analyze and process large volumes of data. This technique allows machines to learn patterns and perform complex tasks, such as speech recognition and computer vision. Its ability to continuously improve as more data is provided to it makes it a key tool in various industries, from health.... From open source the world's largest self-driving dataset to building a 'deep voice artificial intelligence system’ that can clone your voice in seconds, are highly respected in this domain.
And now they have announced their entry into the autoML space with the launch of EZDL. EZDL is a simple drag-and-drop platform that allows users to design and build custom machine learning models. Even if you have no programming experience (and he doesn't have time to learn it either), this tool will feel like a godsend. Its objective is to help those companies that are struggling to expand their artificial intelligence operations and do not have access to a large amount of data (that, frankly, is most of the companies these days).
Nowadays, EZDL offers the option to build two types of custom models:
- Image Models: This includes image classification models and object detection.
- Sound models: This includes models for sound classification and speech recognition..
One of the best features of EZDL is that it only takes four steps to build and implement a model:
- Create a model
- Upload the (s) picture (s) and tag the (the) object (s)
- Train, validate and test the model
- Deploy the Model with Offline SDK Cloud APIs
The company claims it takes less than 15 minutes on average in training models and that more than 2/3 of the models give an accuracy score of more than 90%. A pretty bold claim and one that will undoubtedly be put to the test as the user base grows.
The video below is a nice concise summary of what EZDL does (audio is in Chinese with English subtitles):
Our opinion on this
EZDL offers are still quite limited, but i am impressed with what is shown so far. I tried the object detection service and it was like a breeze. Agree, the image set in my data was small, but it clearly shows that EZDL has potential and Baidu will surely work hard to expand and improve the platform.
AutoML tools are a boon to all non-programmers and the competition in the market makes it easy to try and choose what suits you and your business.. What data scientists think of this tool is another matter, as it completely removes the mathematical aspect of the domain.
