Android camera app that detects the breed of a dog.
Model based in 120 breed of dogs trained over 150 epochs, includes metada and its converted to Tflite to use in android.
Used:
- Python
- TensorFlow
- Keras
- Dataset used for training : Stanford dogs dataset
- Used transfer learning with the MobileNet V2 model
For making predictions with python, keep in mind the model takes 224 x 224 images with 3 RGB Channels.
#here is how i resized it with cv2, you can do it how you prefer
img = cv2.imread(IMAGE_PATH)
resized_img = cv2.resize(img, (224, 224)).astype(np.float32)
final_img = new_img / 255.
output = make_prediction(final_img)Used:
- Kotlin
- MVVM
- CameraX
- Hilt DI
