SpotLab/microfilariae_detection

This model is an object detection model trained with tensorflow object detection API, published with the paper Edge Artificial Intelligence for real-time automatic quantification of filariasis in mobile microscopy

• Developed by: Spotlab

• Model type: SSD mobilenet v2

• Classes: Microfilaria

• Model input: image resized to 640 and normalized to with mean=127.5 and std = 127.5.

• Datasets:

  • Training set: 700 field of view images (100 magnification) from 85 samples with 1965 microfilarias
  • Validation set: 173 field of view images (100 magnification) from 30 samples with 328 microfilarias
  • Test set: 453 field of view images (100 magnification) from 30 samples with 328 microfilarias

• Performance:

  • On validation set: 88.17% precision, 91.62% recall, and 89.85% f1 score.
  • On test set: 94.14% precision, 91.90% recall, and 93.01% f1 score. Example detections The model is trained with the square inside the field of view instead with the full field of view. To ensure the model performace, use use this scale.

You can create your own android app to run this model following this tutorial: TensorFlow Lite Object Detection Android Demo