Parcel and Person Instant Segmentation

Objective

Prepared parcel-and-person segmentation data and experimental outputs for comparing model families.

Stack

Delivery highlights

• Prepared a two-class parcel-and-person dataset by combining media sources and reviewing annotations for instance-segmentation experiments with YOLO and RF-DETR model families.
• Used the asset collection to compare segmentation outputs and support cloud-GPU training experiments. The reviewed root contains dataset and test artifacts; it does not independently verify a reproducible benchmark, identical hyperparameters across every run, or production-ready deployment.
• The next step before publishing results is to document source licences, privacy constraints, dataset splits, training configuration, and per-model accuracy and latency measurements.