Creating a website where all the information, after being analyzed, will be visualized in the form of a dashboard to help visualize the user experience.

• Count the number of 🍾 Cans/Bottles, 📦 Cartons, 👫 People and ♺ POSM; then transform them into distributions as Pie Chart

• Also detect information about Activities and Locations, Atmosphere, Emotion

• Then all of the information, after being processed (drawing bbox, write report,...) will be shown in the end of the website

Core techniques:

• SOTA Prompting Engineering technique: Graph of Thoughts, Chain of Thoughts.

• Details classification using retrieval system: FAISS Vector Space with embedding from SOTA Image Retrieval models (Google/SigLIP, Meta/Dino, Microsoft/BEiT3)  Enhance Accuracy, Scalability comparing to “OCR or using YOLO only” method.

• Training YOLOv10 for classify (Can, Carton, Beer POSM) without depending on the organizer’s training datasets.

• Python: Pytorch, Hugging Face, NumPy, FAISS, FastAPI, etc.
• Website: ReactJS, HTML, CSS, Figma, etc.