How normangopico/prosthetic_yolo26 works
This repository is a YOLO26 training subsystem for a prosthetic-hand research project, not a complete real-time detector: the README describes detection goals, but the tree shows datasets and training code rather than an inference or prosthetic-control entry point. The principal workflow combines object-specific image/label datasets, remaps their classes into a shared dataset, then fine-tunes a pretrained YOLO26 model and saves training artifacts. Separate archived scripts train key and wallet models. Dataset preparation and training-source excerpts were sampled; details beyond those excerpts are represented only where the tree or README supports them.
Main entry points: build_combined_dataset.py Dataset combiner · train_combined.py Combined trainer · solo_training_archives Solo trainers · dataset_combined Combined dataset
Components 19
Object datasets
- Key images
dataset_keys - Wallet images
dataset_wallet - Bottle images
dataset_bottle - Can images
dataset_can - Book images
dataset_book - Phone images
dataset_smartphone - Fork images
dataset_fork - Spoon images
dataset_spoon - Pencil images
dataset_pencil - Apple images
dataset_apple
Dataset preparation
- Dataset combiner
training/build_combined_dataset.py - Combined dataset
dataset_combined
Model training
- Combined trainer
training/train_combined.py - Solo trainers
training/solo_training_archives
Training artifacts
- Combined run
runs/detect/combined_yolo26n - Solo runs
runs/detect
Other components
- Researcher
- Ultralytics YOLO
- Pretrained weights
yolo26n.pt
Connections 20
- Researcher → Dataset combiner: runs preparation
- Key images → Dataset combiner: reads images
- Wallet images → Dataset combiner: reads images
- Bottle images → Dataset combiner: reads images
- Can images → Dataset combiner: reads images
- Book images → Dataset combiner: reads images
- Phone images → Dataset combiner: reads images
- Fork images → Dataset combiner: reads images
- Spoon images → Dataset combiner: reads images
- Pencil images → Dataset combiner: reads images
- Apple images → Dataset combiner: reads images
- Dataset combiner → Combined dataset: writes dataset
- Researcher → Combined trainer: starts training
- Combined dataset → Combined trainer: provides data
- Pretrained weights → Combined trainer: loads weights
- Combined trainer → Ultralytics YOLO: trains model
- Ultralytics YOLO → Combined run: saves run
- Researcher → Solo trainers: starts training
- Solo trainers → Ultralytics YOLO: trains model
- Solo trainers → Solo runs: saves runs