https://github.com/ayyucedemirbas/dermnet_tf_lite_micro
https://github.com/ayyucedemirbas/dermnet_tf_lite_micro
dermnet gsoc-2022 image-classification tensorflow-lite-micro
Last synced: 29 days ago
JSON representation
- Host: GitHub
- URL: https://github.com/ayyucedemirbas/dermnet_tf_lite_micro
- Owner: ayyucedemirbas
- License: gpl-3.0
- Created: 2022-07-11T13:48:44.000Z (almost 4 years ago)
- Default Branch: main
- Last Pushed: 2022-09-16T18:57:46.000Z (almost 4 years ago)
- Last Synced: 2025-03-04T00:55:22.250Z (over 1 year ago)
- Topics: dermnet, gsoc-2022, image-classification, tensorflow-lite-micro
- Language: Jupyter Notebook
- Homepage:
- Size: 981 KB
- Stars: 1
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# dermnet_Tf_lite_micro
## Dataset
The original [Dermnet](https://www.kaggle.com/datasets/shubhamgoel27/dermnet) dataset consists of 23 classes. Since we work with microprocessors that have hardware constraints, we had to reduce the number of classes to 4. The new dataset contains the following classes:
1- Acne and Rosacea Photos
2- Eczema Photos
3- Nail Fungus and other Nail Disease
4- Tinea Ringworm Candidiasis and other Fungal Infections
Also, the original dataset is imbalanced. So, we applied [data augmentation](https://github.com/ayyucedemirbas/vegetables_detection/blob/main/image_data_augmentation.py) techniques manually. We do not prefer to use a Keras layer to do that. Due to it's not a [supported layer](https://github.com/tensorflow/tflite-micro/blob/main/tensorflow/lite/micro/all_ops_resolver.cc) for TensorFlow Lite Micro, we were going to have to build it as a [custom layer](https://www.tensorflow.org/lite/guide/ops_custom).