A non-invasive, pose-aware weight estimation system for live pigs developed as a research project at Iloilo Science and Technology University (ISAT-U). The system uses a standard mobile or fixed camera to capture pig images and applies deep learning and computer vision to estimate weight (kg), height, width, and length — without any physical handling or stress to the animal.

Built as a cross-platform Flutter application (mobile & web) backed by a FastAPI inference server. The CNN multi-output regression model uses EfficientNet-B0 for backend inference and MobileNetV3-Small exported as TFLite for on-device prediction. Pose-awareness is implicitly learned from labeled image data covering pigs in varied natural postures. Video captures are used for frame extraction to augment the training dataset. Designed for practical, low-cost deployment in local Philippine farm and slaughterhouse environments.

• InstitutionIloilo Science and Technology University (ISAT-U)
• ProponentWeena J. Bulnes
• PlatformMobile & Web Application (Flutter)
• StackFlutter, FastAPI, TensorFlow / Keras, EfficientNet-B0, MobileNetV3-Small, TFLite, Python
• ML TaskCNN Multi-Output Regression — Weight, Height, Width, Length
• TypeResearch & Development — Computer Vision / Deep Learning