Malaria detector using caffe and OpenVINO
Sayak Paul
Kolkata, West Bengal
- 0 Collaborators
This project demonstrates an end-to-end parasitized malaria cell detector using caffe and OpenVINO. OpenVINO is specifically is used to optimize the caffe model to make it work on an NCS. ...learn more
Project status: Under Development
Intel Technologies
AI DevCloud / Xeon,
Intel Opt ML/DL Framework,
Movidius NCS,
OpenVINO,
Intel Python
Overview / Usage
Deep learning has shown tremendous progress in medical imaging like many other domains. This project demonstrates an end-to-end malaria detection system from a pool of parasitized and uninfected blood cells. It used caffe to as the deep learning framework and it then uses OpenVINO to optimize the model so that it can be used to run inference on an NCS.
Methodology / Approach
-
I collected the dataset from NIH's official website.
-
Prepared the dataset to make it compatible for use with caffe.
-
Define the network topology and then trained it on DevCloud (took ~23 minutes to train and yielded an accuracy of ~95%).
-
Serialized the model weights in .caffemodel format because OpenVINO accepts it.
-
I then used this serialized model and OpenVINO's Model Optimizer to generate the IR files.
-
Used the IR files with Inference Engine to run inference on an NCS.
Technologies Used
-
Python
-
caffe
-
OpenVINO
-
DevCloud