The job of Machine learning Team is to learn and make machine learning model of pink eye and healthy eye of goat and cow. We make 2 model for this App, 1 model is for cow that contain 2 class, pinkeye and healthy. And 1 model is for goat that also contain 2 class, pinkeye and healthy.
We have to read journal or other resource from internet to gain more knowledge about the pinkeye disease that affect cattles.
We get information from various source such as:
- Mengenal Penyakit Pink Eye (Mata Merah) Pada Ternak Kambing & Domba
- Pink Eye Cases in Goats at The Sawangan Farm
- Penyakit Pinkeye Pada Sapi
- PENGOBATAN DAN PENCEGAHAN PENYAKIT PINK EYE PADA TERNAK
We collect images of pink eye and healthy eye from cow eye and goat eye
We get dataset from various source such as:
After collecting the datasets, we do preparation data by cleaning the images we found not suitable for our model, such as deleting and crop image that still can be use.
We use image augmentation by applying various transformations to the original images to creates additional training samples with different variations. Image augmentation is useful to expanding the training data with diverse variations and to reduce overfitting.
For model architecture, resize the image into 180x180 and we use transfer learning using MobilenetV2.
In general, the Architecture of MobileNetV2:
Cow distribution dataset
Cow model accuracy and loss
Goat model accuracy and loss
Goat model accuracy and loss
This is the example ipynb of how we test the images to classify the eye using the model we created before and here is the result: Testing
- This project is created for educational purpose as the requirement to graduate from Bangkit Academy led by Google, Tokopedia, Gojek, & Traveloka.