Efficient Convolution Neural Network Software Implementation and DS Preprocessing for AIRBiS
AIRBiS Overview†
Background†
With the Coronavirus disease (COVID-19) disrupting large part of the world, the number of patients has grown progressively. For dealing with this serious emergency, accurate diagnosis and fast reporting are two significant mechanisms.
Research Goal†
- Search and collect X-ray image datasets containing non-infected (normal) and infected (COVID-19) data given many open sources
- Perform pre-processing tasks, e.g. as follows:
- denoising
- resizing
- data augmentation
- segmentation
- we should also come up with other issues and corresponding strategies
- CNN implementation and evaluation
- A CNN-based architecture will be adopted for the binary classification task. Once the model learns the feature representation of the input images, a classifier can thus output the diagnosis result, i.e., normal or infected. Framework and language: TensorFlow (Keras), Python
- Using the diagnosis (classification) results to evaluate the performance of the CNN (e.g., detection accuracy)
Research Schedule†
| Date | Task |
| June 1 - July 15, 2020 | Making prototype CNN |
| July 15 - Aug 15, 2020 | Preprocessing the Dataset |
| Aug 15 - Aug 31, 2020 | Argumentation the Dataset |
| Sep 1 - Sep 31, 2020 | Optimization CNN |
| Oct 1 - Dec 1, 2020 | Writing a paper |
| Deadline for 1st Draft: 1st December 2020 | International ACM conference |
Research Progress†
References†
![[PukiWiki]](image/pukiwiki.png "[PukiWiki]")
