Folders and files
Image Classifier App is a Command Line Application that build for AI Programming with Python Nanodegree Program that provided by udacity. The main feature of the app is to classifiy Image using pretrained model like vgg16.
The library the has been used to develop this app.
| Library | Link |
|---|---|
| pytorch | https://pytorch.org/ |
| matplotlib | https://matplotlib.org/ |
| numpy | https://numpy.org/ |
| json | https://docs.python.org/3/library/json.html |
| argparse | https://docs.python.org/3/library/argparse.html |
| torchvision | https://pytorch.org/docs/stable/torchvision/index.html |
| PIL | https://pillow.readthedocs.io/en/stable/ |
| time | https://www.programiz.com/python-programming/time |
python3 train.py data_dir --learning_rate LEARNING_RATE --epochs EPOCHS --save_dir SAVE_DIR
example for this specific parameters:
- data_dir : "flowers_10"
- learning_rate : 0.0001
- epochs : 6
- save_dir : checkpoint.pth
python3 train.py flowers_10 --learning_rate 0.0001 --epochs 6 --save_dir checkpoint.pth
Argument Info data_dir : flowers_10 save_dir : checkpoint.pth learning_rate : 0.0001 epochs : 6 Start Training Epoch: 1/6.. Training Loss: 6.643.. Test Loss: 5.841.. Test Accuracy: 0.116 Epoch: 1/6.. Training Loss: 5.228.. Test Loss: 3.421.. Test Accuracy: 0.116 Epoch: 2/6.. Training Loss: 3.153.. Test Loss: 2.428.. Test Accuracy: 0.256 Epoch: 2/6.. Training Loss: 2.603.. Test Loss: 2.281.. Test Accuracy: 0.279 Epoch: 3/6.. Training Loss: 2.460.. Test Loss: 1.440.. Test Accuracy: 0.605 Epoch: 3/6.. Training Loss: 1.881.. Test Loss: 1.309.. Test Accuracy: 0.512 Epoch: 3/6.. Training Loss: 1.851.. Test Loss: 1.119.. Test Accuracy: 0.558 Epoch: 4/6.. Training Loss: 1.470.. Test Loss: 0.931.. Test Accuracy: 0.698 Epoch: 4/6.. Training Loss: 1.170.. Test Loss: 0.873.. Test Accuracy: 0.651 Epoch: 5/6.. Training Loss: 1.024.. Test Loss: 0.784.. Test Accuracy: 0.651 Epoch: 5/6.. Training Loss: 0.807.. Test Loss: 0.565.. Test Accuracy: 0.837 Epoch: 6/6.. Training Loss: 0.754.. Test Loss: 0.461.. Test Accuracy: 0.837 Epoch: 6/6.. Training Loss: 0.690.. Test Loss: 0.385.. Test Accuracy: 0.884 Epoch: 6/6.. Training Loss: 0.468.. Test Loss: 0.360.. Test Accuracy: 0.907 End Training The total training time: 42.188 minuates Saved to checkpoint.pth
python3 predict.py image_path checkpoint --top_k TOP_K --category_names CATEGORY_NAMES
example for this specific parameters:
- image_path : "flowers_predict/image_06763.jpg"
- checkpoint : checkpoint.pth
- top k : 10
- category names : cat_to_name_10.json
python3 predict.py flowers_predict/image_06763.jpg checkpoint.pth --top_k 10 --category_names cat_to_name_10.json
Argument Info
image_path : flowers_predict/image_06763.jpg
checkpoint : /Users/Apple/Documents/python_IA/Image_Classifier/train/checkpoint.pth
top k : 10
category names : cat_to_name_10.json
Predict Result
('pink primrose', 0.99950004)
('sweet pea', 0.00045578534)
('canterbury bells', 2.4515823e-05)
('moon orchid', 1.9176043e-05)
('monkshood', 2.6036224e-07)
('english marigold', 1.32129e-07)
('globe thistle', 5.6247494e-08)
('hard-leaved pocket orchid', 7.674287e-09)
('tiger lily', 9.374542e-10)
('bird of paradise', 4.8567556e-10)
MIT