I was assigned to the Personal Colour Analysis team 😁. I implemented the pipeline from the customer face taken to output the customer's skin tone utilizing Python!🐍. I use Google Face Landmark to extract some of the features from the face to be used, and I implemented the white balance to correctly extract the colors. There are 4 skin tones: warm, cool, summer, and winter, using the math formula and combining with fuzzy logic✨.

I am a bit of a designer too!🎨. I had the chance to design the banner for the live collaboration used in Matahari Department Store

Face preprocessing is utterly complicated 😓, when I built the pipeline for the Personal Color Analysis it involved in 4 steps

• Customer face pose for the camera
• Google Face Landmark extracted important parts
• White balance Implementation
• Math formula and fuzzy logic

Steps 2-3 were not the problems; the 1st and 4th were. First, we need to ensure that the lighting is not too dark 🌑 or too bright ☀️ because it will affect how the WB works. The position also depends on how far the customer faces from the camera.

The fourth step is more "good" theoretically on paper 📃 rather than in implementation. Most of the skin tone still needs improvement, especially in summer and winter. We could improve if we had more data, time, and research ⌛.

I was having fun doing this project, I learned more about Computer Vision 🙂 pipelining, and understood how certain areas of our faces 👨‍🦰 are capable of matching our outfits!🧢.