Review of last issue
In the last issue, we used CTC technology to train and recognize blue cards and license plates in the same network, and achieved good results. This issue is mainly to carry out relevant statistics on the road, on the one hand to test the performance of the program, on the other hand to see how much new energy vehicles account for. The result is really beyond my expectation.
Select a statistical location
In order to ensure the richness of statistical sampling, four places were selected: bustling urban road sections, remote road sections, open-air parking lots, and underground parking garages. Open the map of "International Bin" (Binjiang District, Hangzhou City), compare the above four locations, and make a selection. We chose the underground parking garage of a shopping mall in the west of the city, the Xianghu parking lot in the south of the city, the expressway in the north of the city (busy road section), and Beitanghe Road in the east of the city (remote road section).
statistics video
The animation below shows the video at the time of statistics. Due to size limitation, it can only be displayed for about 3 seconds
statistical results
The final statistical results are as follows:
- Underground parking garage: fuel vehicles: 112; new energy vehicles: 27; proportion of new energy vehicles: 19.4%
- Open-air parking lot: fuel vehicles: 219; new energy vehicles: 65; proportion of new energy vehicles: 22.3%
- Bustling roads: fuel vehicles: 1016; new energy vehicles: 191; proportion of new energy vehicles: 15.8%
- Remote roads: fuel vehicles: 315; new energy vehicles: 77; proportion of new energy vehicles: 19.6%
Weighted average of the above results, the final proportion of new energy vehicles: 17.8%. The result was quite unexpected. Although I know that there must be no half, but I think there should be about 30%, but through actual tests, I found that it is less than 20%. Seeing this result, I suddenly remembered this picture. Back then, Li Bin made an astonishing remark, mocking that people who bought fuel vehicles were out of nostalgia. I feel that the salesman was brainwashed by him!
Respond to sales
With the above data, I am confidently ready to go back to sales (if you don’t know the storyline, please read the first issue). After I package all kinds of data, I send it to sales. After a month, I don’t know if she still forgot the nonsense she said back then (she told me that people who buy petrol cars now are fools, and that more than half of the cars on the road are new energy cars). I imagined that after being slapped in the face by my data, she could lower her proud head in front of me and give her eyes of admiration. . . However, the real situation is this:
Ten thousand mud horses rose in my heart. . . This taste is indescribable! His blood pressure soared, and he was in the process of pinching someone to save himself. . . After quite a while, the mood slowly returned to calm. But looking back, although I didn't succeed in sales, I learned a lot in the process, completed the practice, and helped some friends with the same hobbies, which is also a great comfort.
Summarize
The span of this project is relatively long, with a total of 5 phases. The cause, network design, problems encountered and corresponding improvements are fully and detailedly explained. I hope it will be helpful to those friends who want to do some small deep learning projects. I will also open source the source code of this project. However, it is still in the process of sorting out, please keep an eye on it, and notify you as soon as it is open source. Finally, once again, I would like to thank everyone for accompanying me to the last episode. I wanted to give up several times in the middle. It was the support of fans that made me persevere, thank you.