amaghini/tensorflow-image-classification-retrainable repository overview

⁠Retrainable Tensorflow image classification model - Docker based

run docker build . -t amaghini/tensorflow-image-classification-retrainable:latest

---

Once image is built you can

docker run --env-file .env --mount src="$(pwd)/__tests__",target=/tf/__tests__,type=bind --mount src="$(pwd)/input_photos",target=/tf/photos,type=bind --mount src="$(pwd)/tf_files/output",target=/tf/tf_files/output,type=bind amaghini/tensorflow-image-classification-retrainable:latest

to see the classification result for the test_input.jpg file

---

Once trained, if you want to see the classification result for a different image you can

docker run --mount src="$(pwd)/__tests__",target=/tf/__tests__,type=bind --mount src="$(pwd)/input_photos",target=/tf/photos,type=bind --mount src="$(pwd)/tf_files/output",target=/tf/tf_files/output,type=bind amaghini/tensorflow-image-classification-retrainable:latest python scripts/label_image.py --graph=tf_files/output/retrained_graph.pb --labels=tf_files/output/retrained_labels.txt --input_layer=Placeholder --output_layer=final_result --image=__tests__/$TEST_FILE

---

If you want to access the container, you can simply

docker run -it amaghini/tensorflow-image-classification-retrainable:latest bash

---

⁠Notes

• retrain.py script coming from https://raw.githubusercontent.com/tensorflow/hub/master/examples/image_retraining/retrain.py⁠
• label.py script coming from https://raw.githubusercontent.com/tensorflow/tensorflow/master/tensorflow/examples/label_image/label_image.py⁠