Livoa LogoLivoa
Input Stage


Raw Waste Images (Kaggle, Medical Datasets)

Proofreading (Rescan, Medical Datasets)

Processing: Resize, Normalize, Augment

Encoder: Feature Extraction


ResNet50 Backbone Waste

YOLOv5/v8 Backbone (Mixed Waste)

Medical Model (Synogure Maps, Spatial + Semantics, etc.)

Loss Function (Cross-entropy, IoU)
Bottleneck: Latent Representation Features
Optimizer (Adam) + Backpropagation
Decoder: Classification & Detection


General Waste → Classes:

Medical, Plastic, Glass, Paper, Metal

Item Type + Color-Coded Eco Bin

Training Loop


Iterate until Convergence (Early Stopping)

Output Stage: Disposal Guidance + Rewards


Waste Type, Bin Color, Credit Awards, EcoPoints

AI Model


ResNet50 Backbone Waste

Medical Waste

Item Type + Color-Coded Bin

User Upload


(Web/Mobile Frontend - React/Flutter)

Flask API


(Python Backend)

MongoDB Database


User History, EcoPoints

RECIRCLE

by SAISH

0
0 uses