In-depth analysis of the five dimensions, the entry AI first needs to choose the direction

The pace of human use of machines to help production has never stopped. To enter the AI ??field, we must first understand the current structural system of the artificial intelligence industry:

“Basic support” and “commercial scenario” are issues at the enterprise level. For personal development, it is necessary to upgrade at the “core technology” level – it is king to have the workplace ability of the AI ??industry!

If you want to step into the AI ??field, what direction should you choose?

Take the investment data of each track in the AI ??field in 2017 for an analysis:

Overall, the largest number of investment events is computer vision, followed by natural language processing, intelligent robots and autonomous driving. At the same time, according to other data, computer vision has the largest number of startups in the field of artificial intelligence, accounting for 17.7%.

Source: 2017 White Paper on the Development of Artificial Intelligence Industry

So what does artificial intelligence-computer vision specifically do?

Computer vision refers to the use of machines to simulate "visual organs", to identify, track and measure targets, and to replace the brain with computers for further image processing and interpretation. At present, the face recognition unlocking commonly used in our mobile phones, and the remote processing of banking services are all used in this type of technology.

So what is the current environment of computer vision?

I have collected job recruitment information from multiple cities for nearly one month from a recruitment website. We use data to interpret:

If the position salary is 20k-40k: the minimum value is 20k, the maximum value is 40k, and the average value is 30k. The blue line in the figure is the median value.

From the "mean" indicator, we can see that the monthly salary of computer vision direction is about 30k. If the senior algorithmist can get a monthly salary of more than 40k, the "minimum" indicator can be seen that the starting salary of the position is also 20k. .

Here, several cities are extracted to see the average monthly salary. The first-tier cities are basically maintained at 24k-30k, and the second-tier cities are basically above 15k.

Among the existing data, the industry distribution of posts is mostly in the “mobile Internet”.

According to iiMedia's data report, the most popular use of Chinese Internet users in 2017 is also related to smartphone terminals. Be aware that "picture beautification", "face recognition application", "smart album", etc. all fall into this category.

From the data, we can also see keywords such as “e-commerce”, “finance”, “social network” and “health health”.

In 2017, computer vision companies received huge financing, such as domestic head enterprises Shangtang Technology, Vision Technology, and Yun Cong Technology (both have reached hundreds of millions of financing); at the same time, from the company financing data , you can see that the computer Most of the visual jobs are concentrated in startups and mature listed companies, with a good corporate environment.

In terms of job qualifications, master's degree is the largest, followed by undergraduate, and doctors account for a small proportion. The latter may be affected by the scarcity of high-quality talents.

At present, the undergraduate education in China is not as good as the master's, and the undergraduate's plasticity is strong. It is also a good personal development route to enter the enterprise early.

I have already placed the data of the above recruitment website in the discussion group at the end of the text. In order to make up for the lack of data, I also compiled some official industry reports for the reference of the small partners. Please enter the group to find in the group file.

So if Xiaobai wants to choose a direction to cut into the AI ??field, it is highly recommended to choose computer vision! This is a direction with huge development potential, mature business scenarios, and shortage of talent demand.

But since this salary is so high, why are there fewer people who are engaged in this position than you think? This raises the next question -

Learn from scratch, you need to turn over several mountains!

For a small partner with a complete base of 0, if you want to enter the AI ??field, there are four mountains that you need to climb:

You need to learn a programming language

At present, the mainstream artificial intelligence algorithms need to be implemented on the basis of code, and the related toolkits are called. Here, I recommend Python, the main development language of artificial intelligence.

Python is a glue language with a rich and powerful library. It also has strong support for deep learning. The most important thing is that it is very friendly to Xiaobai. It is easy to learn. It is no exaggeration to say that Python is the future of AI and machine learning. .

You need to "master" advanced mathematics

The mathematical foundation is probably to scare off most people's paper tigers. From calculus to linear algebra, we really need to understand the mathematical principles. But this does not mean that you have to do the same calculations as the university's high numbers, and do some exam questions that you don't even know why. The core of the artificial intelligence mathematics part is to use code to achieve, understand the deep logic of the data formula, and skillfully apply the code. Python's rich and powerful libraries will help you a lot.

You need to understand the algorithm - deep learning (neural network)

When you use code power and mathematics, the next step is to understand the core algorithm of artificial intelligence - deep learning. Deep learning is a subclass of machine learning and a technology for machine learning: training deep neural networks with supervised or unsupervised learning methods to make computers "smart."

In the computer vision direction learning, it is necessary to analyze the image data and train the algorithm model to solve the tasks of detection, segmentation and recognition.

You need to use practice to test what you have learned.

After you have crossed the first few mountains, you have mastered the basic skills of AI systematically. The next step is to understand the pain points of actual work.

To put it simply, although we can call Python deep learning tools such as Tensorflow and Keras, when solving practical problems, we need to understand the classic algorithms summarized by the seniors and learn to optimize the model. The complexity of this stage is very complicated. High, it takes a lot of time to lick and digest.

City Data Group + NetEase Jointly Launches Artificial Intelligence Course

After more than three months of preparation and design, Netease and the City Data Group jointly launched the "AI Engineer (Computer Vision)" micro-professional course.

This course will start from Python basics, to mathematical theory knowledge, and deep learning algorithms. Finally, through 15 actual combat projects, use 3 months to get zero-based white introduction AI!

In the Python language part we will start with the basic syntax and then go to the core toolkit for data analysis: numpy, pandas, matplotlib, and the computer vision library OpenCV;

In the mathematics section, you will include the most basic functions of higher mathematics, to calculus to linear algebra, and probability theory, while teaching you to use Python to implement complex formulas, deepen understanding, and then step in;

Deep learning algorithm level will take

You build a simple algorithm framework, such as single-layer perceptron, BP neural network, and then learn to call Tensorflow, Keras toolkit to build the framework.