Project Overview
Logistic systems are slowly moving on to using AI as their aid to differentiate shipping equipment across the board. When it comes to the aiding, tracking, and monitoring process. AI has introduced large advances that make it easier for workers to process shipping data accordingly.
At the request of a logistics client, I was tasked with designing 3-D shipping containers for an AI/ML project that parsed shipping data, and details based on the parameters demanded.
The deliverable produced was a dataset with pre-requisite code that allowed the ML engineers to train their models accordingly to solve the problem.
My Contribution
To design the initial models, I used Blender 2.9 for this project. After rendering all the assets, I used a python script to automate the color and serial on every container.
After every "8 frames" upon the development specifications of my client.
Project overview
To design the initial models, I used Blender 2.9 for this project. After rendering all the assets, I used a python script to automate the color and serial on every container, after every "8 frames" upon the development specifications of my client.
In the beginning of script i store current frame of render in an variable which tell the script which frame its currently rendering below:
Serial Number Manager
This code focuses on change serial number on container at time of render after every 8th frame till end frame it change serial number on contianer.
Node Configuration for Colour & Serial Number Changer
I used the time info node to get the current time frame at the time of rendering. And the connected output of that node to the math node that divided the time frame value by 8. So after every 8th time frame of render, the color and the serial number of the container will change upon request.
Thus I used another node that converts float types to integer, as it is not a good practice when using floating-point values over integer types when it comes to these projects.
​
Afterward, I created another script node that loaded my main script containing data about serial numbers. Alongside this node, I used two text output nodes that displayed the container details on the front and rear end of my container, which would change on every 8th frame of the render.
​
​
Conclusion
This project taught me industry practices when it comes to optimizing 3-D models for AI/ML datasets. In the future, I will try using better poly counts that optimize the size of the data models for the required ML activities.