image handle tool and machine learning algorithm.

• Based on complete training Cell Detector Model, we train it with a small amount other type cell Date.
• And finish detector task with relatively High Accuracy (Compare to other commercial cell detector Like Countess).
• We have more than 80% confidence level

keremberke/yolov8m-blood-cell-detection

Image Tool

• Split the cell image into 540 X 540
• Apply thresholding using Otsu's method (with Gaussian blur pretreatment)
• Find contours in the thresholded image
• Invert contours to Label Studio format
• Update labeling semi-automatically

Machine Learning

• Training blood-cell-detection Model with Labeling date
• Finish Detector Task