Skip to content

Latest commit

 

History

History
32 lines (17 loc) · 1.28 KB

README.md

File metadata and controls

32 lines (17 loc) · 1.28 KB
Raw

NuclearLobeDetection

Outputs

Inference

Inference

  • Example image with bounding boxes drawn around the detected nuclear lobes. Showing the classification and confidence score of each detection.

Genetic Algorithm (Hypertuning of YOLO model)

Box Plot showing the distribution of fitness scores per generation

FitnessDistributionPerGeneration

  • Fitness is determined using metrics/mAP50-95(B) (mean average precision over 50-95% IoU (Intersection over Union) thresholds) and mAP50 (mean average precision over 50% IoU threshold) for the validation set.
  • This means that the fitness score is a value between 0 and 1, where 1 is the best possible score.

Line Graph showing the trend of the average fitness score per generation

AverageFitnessPerGeneration

Technical Documentation

YOLO

For each image in images/train or images/val, YOLO expects a corresponding .txt file with the same name in the labels/train or labels/val directory.

Genetic Algorithm

Largely based on my UG Diss: https://github.com/danielmusselwhite/UG_Dissertation_Energy_Forecasting/blob/main/Src/Jupyter_Notebooks/4.%20LGBM%20forecasting%20Genetic%20Algorithm.ipynb