PollEncOuNt 0.0.6

The purpose of this project is to count the number of alive and dead pollens in an image.

PollEncOuNt

The purpose of this project is to count the number of alive and dead pollens in an image.

Contents

• Installation

• Outputs Demonstration

• Usage

  • GUI

  • Python API

Installation

1. Package installs

pip install PollEncOuNt

Outputs Demonstration

Raw	Predicted

blue squares detected are alive, green squares detected ard dead.

count_result.csv summarizes the results after running the prediction process.

Filename	Alive	Dead
pollen_predicted.jpg	96	20

Usage

GUI

Launching the GUI

Open your terminal (Mac/Linux) or command line (Windows).

Launch the GUI by typing:

% peon

The PEON Main Menu will appear.

Main Menu

From the Actions menu, you can:

1. Select Train to open the Train GUI.

2. Select Predict to open the Predict GUI.

3. Exit the application.

Predict GUI

A Pre-trained Model can directly be used for prediction.

1. Open the Predict GUI:

• From the main menu, go to Actions > Predict.

2. Configure Prediction Settings:

• Image Files: Click the Browse button and select one or more image files for prediction.
• Model File: Click the Browse button to load the trained model file (.pt, .pth, or .onnx).
• Project Save Directory: Click the Browse button to select the directory where the prediction results will be saved.
• Pridect Settings:
  • Check the Save Images box to save the images with predictions.
  • Check the Save CSV box to save the results in a count_result.csv file.
  • Set the Confidence threshold (0-1) to filter detection results. Higher values result in fewer but more confident detections.
  • Set the Int./Union (IoU) threshold (0-1) used for non-maximum suppression. Higher values allow more overlapping bounding boxes (default: 0.5).

3. Start Prediction:

• Click the Start Prediction button to begin inference.
• Real-time log updates will appear in the LOGS section.

4. Reset Prediction:

• Click the Reset button to clear inputs and logs.

Train GUI

1. Open the Train GUI:

• From the main menu, go to Actions > Train.

2. Configure Training Settings:

• Data YAML File: Click the Browse button and select a valid YAML file for training data.
• Project Save Directory: Click the Browse button and choose the directory where training outputs (e.g., logs, model checkpoints) will be saved.
• Model Selection:
  • For Pre-trained Models: Click the Browse button to select a .pt, .pth, or .onnx model file.
  • For YOLO Models: Use the dropdown menu to choose one of the YOLO models (e.g., yolov8n.pt).
• Training Settings:
  • Set the number of Epochs for training.
  • Select the device (cpu or gpu) from the dropdown menu.

3. Start Training:

• Click the Start Training button to begin.
• Monitor the real-time log updates in the LOGS section.

4. Reset Training:

• To reset the form, click the Reset button. This clears all input fields and logs.

Python API

Predict

A pre-trained model can directly be used for prediction(path: ./model/best.pt).

from peon import peon_predict

results = peon_predict(
    img_files=["IMG1", "IMG2", ...],
    model_path="MODEL_PATH",
    save_dir="SAVE_DIR",
    save_img=True,
    save_csv=True,
    conf_thres=0.2,
    iou_thres=0.5,
)

img_files: list[str]

• List of image path to conduct prediction.

model_path: str:

• Path to the YOLO model.

save_dir: str

• Path to the directory to save the results. Required if save_img or save_csv is True.

save_img: bool (optional, defaults True)

• Whether to save the predicted images.

save_csv: bool (optional, defaults True)

• Whether to save the results to a CSV file.

conf_thres: float (optional, defaults 0.2)

• Confidence threshold for filtering detections. Range 0-1. Higher values result in fewer but more confident detections.

iou_thres: float (optional, defaults 0.5)

• Intersection over Union (IoU) threshold for non-maximum suppression. Range 0-1. Higher values allow more overlapping bounding boxes.

Train

from peon import peon_train

peon_train(
    data_path="DATA_PATH",
    save_dir="SAVE_DIR",
    model_path = "yolov8m.pt",
    epochs=100,
    device="cpu",
)

data_path: str

• Path to the data .YAML file.

save_dir: str

• Path to the directory to save the trained model.

model_path: str (optional, defaults "yolov8m.pt")

• Path to the YOLO model.

epochs: int (optional, defaults 100)

• Number of epochs for training.

device: str (optional, defaults "cpu")

• Device to use for training.