ML workflow to detect NOW on sticky traps using scinet and yolo.

python3 -m venv venv_yolo
source venv_yolo/bin/activate
pip install -r requirements.txt

pip install ipykernel
python -m ipykernel install --user --name=venv_yolo --display-name "YOLO (venv_yolo)"

train: path/to/train/images
val: path/to/val/images
names:
  - class1
  - class2

from ultralytics import YOLO

model = YOLO("yolov8n.pt")
model.train(
    data="your_data.yaml",
    epochs=25,
    imgsz=640,
    batch=8,
    project="my_project",
    name="my_experiment"
)

• organise.py: Arranges dataset into folders for traning
• rotate.py: Rotates images
• fix.py: Fixes broken annotations or paths
• format_and_separate.py: Moves and formats files by type
• duplicate.py: Detects and filters out duplicate files
• voc.py: Converts annotations to/from Pascal VOC format
• rm-file.txt: List of files to delete from dataset not needed for most applications

• train.py: Main training runner
• train_now_6_25_2025.py: Specific experiment training script
• modelTest.py: Runs predictions and outputs images with bounding boxes
• model_analytics_demo.py: Loads trained model and shows visual metrics
• model_analytics_function.py: Helper functions for plotting precision, recall, confidence, and loss
• compare_model.py:Graphs confidence scores for 2 models
• count-export.pyput in folder to script and get out NOW counts csv

• demo.py: Fast sample inference runner
• accurate.py: Compares predictions against ground truth
• bet.py: May experiment with detection thresholds or model output settings
• count.py: Tallies number of detections per image or set

• data-template.yml: Template YOLO data config
• job.sh: SLURM shell script to run training on a cluster not needed for most
• make-venv: Bash script to set up Python virtual environment
• setup.txt: Extra setup notes and instructions