Upload 11 files

Procfile

@@ -0,0 +1 @@
+web: sh setup.sh && streamlit run app.py

README.md

@@ -1,13 +1,17 @@
----
-title: Poizon App
-emoji: 📉
-colorFrom: green
-colorTo: green
-sdk: streamlit
-sdk_version: 1.21.0
-app_file: app.py
-pinned: false
-license: mit
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# Streamlit app for trying out poison oak classifier
+Try it out: https://poison-app.herokuapp.com (note: may take ~30sec for webpage to load since the server is on a non-paid licens)
+
+## What is it
+This is a image classifier trained on hand-collected poison oak images over 1 year. Poison ivy will not necessarily expected to be classified as poison oak even though it's leaves also contain the oil causing the allergic reaction, Sumac.
+
+At time of repository creation, this model was implemented in a free iOS app "Poizon Plants", however due to Apple developer licence costs it may not be live.
+
+In order to help the user assess interpret the probability of a given image containing poison oak, different prompts based on the softmax probability are output.
+
+Model is initialized at same time as server start-up to speed up inference. (Tensorflow lite was tested and optimized for latency, but did not significantly improve speed but did significantly decrease f1 metrics)
+
+## Run
+*streamlit run app.py*
+
+## Tests
+pytest

app.py

@@ -0,0 +1,39 @@
+import streamlit as st
+from PIL import Image, ImageOps
+import io
+from img_classification import teachable_machine_classification, load_model
+
+from tensorflow import keras
+
+
+st.set_option("deprecation.showfileUploaderEncoding", False)
+
+st.title("Detecting presence of Poison Oak")
+st.header("Poison Oak Classification Example")
+st.text("Upload an image for classification as poison oak or no poison oak")
+
+
+# Load trained model
+model = load_model("./best_model.h5")
+
+print("Starting Streamlit app")
+uploaded_file = st.file_uploader("Select an image ...", type=["jpg", "png", "jpeg"])
+
+if uploaded_file is not None:
+    image = Image.open(uploaded_file)
+    st.image(image, caption="Uploaded image", use_column_width=True)
+    st.write("")
+    st.write("Classifying...")
+    label = teachable_machine_classification(img=image, model=model)
+    if label <= 0.2:
+        st.write("Very unlikely that this is poison oak.")
+    elif (label > 0.2) & (label <= 0.6):
+        st.write(
+            "Unsure from this picture. You may need to retake a closer/clearer picture."
+        )
+    elif (label > 0.6) & (label <= 0.7):
+        st.write("Decent chance that this is poison oak.")
+    else:
+        st.write("{:.1f}% chance that this might be poison oak".format(label * 100))
+else:
+    st.write("No file uploaded")

best_model.h5

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:148242b744c1e8115f7a417fae172b938b8d9ebb9e387b2690dbd478f1e6ea5b
+size 6860424

img_classification.py

@@ -0,0 +1,63 @@
+from tensorflow import keras
+from PIL import Image, ImageOps
+import numpy as np
+import io, os
+import logging
+import keras_metrics
+from tensorflow import keras
+import utils
+## Configs
+keras.utils.get_custom_objects()['recall'] = utils.recall
+keras.utils.get_custom_objects()['precision'] = utils.precision
+keras.utils.get_custom_objects()['f1'] = utils.f1
+
+
+def teachable_machine_classification(img=None, model=None):
+    """Performs inference on image uploaded"""
+
+    # Create the array of the right shape to feed into the keras model
+    data = np.ndarray(shape=(1, 299, 299, 3), dtype=np.float32)
+    image = img
+    # image sizing
+    size = (299, 299)
+    image = ImageOps.fit(image, size, Image.ANTIALIAS)
+
+    # turn the image into a numpy array
+    image_array = np.asarray(image)
+
+    # Normalize the image
+    normalized_image_array = (image_array.astype(np.float32) / 127.0) - 1
+
+    # Load the image into the array
+    data[0] = normalized_image_array
+
+    # run the inference
+    prediction = model.predict(data)
+
+    print("Prediction", prediction)
+    return prediction[0][
+        1
+    ]  # np.argmax(prediction) # return position of the highest probability
+
+
+def load_model(weights_file=None):
+    """Loads trained keras model"""
+    dependencies = {
+        "binary_f1_score": keras_metrics.binary_f1_score,
+        "binary_precision": keras_metrics.binary_precision,
+        "binary_recall": keras_metrics.binary_recall,
+    }
+
+    try:
+        assert os.path.exists(weights_file), f"File '{weights_file}' does not exist"
+        # Load the model
+        model = keras.models.load_model(
+            weights_file, custom_objects=dependencies, compile=False
+        )
+
+        return model
+    except Exception as e:
+        logging.error("ERROR: ", e)
+        print("ERROR: ", e, " Failed to load ML model")
+
+        return None

requirements.txt

@@ -0,0 +1,114 @@
+absl-py==0.11.0
+altair==4.1.0
+appnope==0.1.0
+argon2-cffi==20.1.0
+astor==0.8.1
+astroid==2.4.2
+astunparse==1.6.3
+async-generator==1.10
+attrs==20.3.0
+backcall==0.2.0
+base58==2.0.1
+bleach==3.2.1
+blinker==1.4
+boto3==1.16.12
+botocore==1.19.12
+cachetools==4.1.1
+certifi==2020.6.20
+cffi==1.14.3
+chardet==3.0.4
+click==7.1.2
+decorator==4.4.2
+defusedxml==0.6.0
+entrypoints==0.3
+enum-compat==0.0.3
+gast==0.3.3
+gitdb==4.0.5
+GitPython==3.1.11
+google-auth==1.23.0
+google-auth-oauthlib==0.4.2
+google-pasta==0.2.0
+grpcio==1.33.2
+h5py==2.10.0
+idna==2.10
+ipykernel==5.3.4
+ipython-genutils==0.2.0
+ipywidgets==7.5.1
+isort==5.6.4
+jedi==0.17.2
+Jinja2==2.11.2
+jmespath==0.10.0
+jsonschema==3.2.0
+jupyter-client==6.1.7
+jupyter-core==4.6.3
+jupyterlab-pygments==0.1.2
+Keras==2.4.3
+keras-metrics==1.1.0
+Keras-Preprocessing==1.1.2
+lazy-object-proxy==1.4.3
+Markdown==3.3.3
+MarkupSafe==1.1.1
+mccabe==0.6.1
+mistune==0.8.4
+nbclient==0.5.1
+nbconvert==6.0.7
+nbformat==5.0.8
+nest-asyncio==1.4.2
+notebook==6.1.5
+numpy==1.18.5
+oauthlib==3.1.0
+opt-einsum==3.3.0
+packaging==20.4
+pandas==1.1.4
+pandocfilters==1.4.3
+parso==0.7.1
+pathtools==0.1.2
+pexpect==4.8.0
+pickleshare==0.7.5
+Pillow==8.0.1
+prometheus-client==0.8.0
+prompt-toolkit==3.0.8
+protobuf==3.13.0
+ptyprocess==0.6.0
+pyarrow==2.0.0
+pyasn1==0.4.8
+pyasn1-modules==0.2.8
+pycparser==2.20
+pydeck==0.5.0
+Pygments==2.7.2
+pylint==2.6.0
+pyparsing==2.4.7
+pyrsistent==0.17.3
+python-dateutil==2.8.1
+pytz==2020.4
+PyYAML==5.3.1
+pyzmq==19.0.2
+requests==2.24.0
+requests-oauthlib==1.3.0
+rsa==4.6
+s3transfer==0.3.3
+scipy==1.5.4
+Send2Trash==1.5.0
+six==1.15.0
+smmap==3.0.4
+streamlit==0.70.0
+tensorboard==2.3.0
+tensorboard-plugin-wit==1.7.0
+tensorflow-cpu==2.3.1
+tensorflow-estimator==2.3.0
+termcolor==1.1.0
+terminado==0.9.1
+testpath==0.4.4
+toml==0.10.2
+toolz==0.11.1
+tornado==6.1
+traitlets
+tzlocal==2.1
+urllib3==1.25.11
+validators==0.18.1
+watchdog==0.10.3
+wcwidth==0.2.5
+webencodings==0.5.1
+Werkzeug==1.0.1
+widgetsnbextension==3.5.1
+wrapt==1.12.1

runtime.txt

@@ -0,0 +1 @@
+python-3.8.6

setup.sh

@@ -0,0 +1,13 @@
+mkdir -p ~/.streamlit/
+
+echo "\
+[general]\n\
+email = \"[email protected]\"\n\
+" > ~/.streamlit/credentials.toml
+
+echo "\
+[server]\n\
+headless = true\n\
+enableCORS=false\n\
+port = $PORT\n\
+" > ~/.streamlit/config.toml

tests/test_static.py

@@ -0,0 +1,14 @@
+###############################################
+# Tests for existence of files not in remote
+###############################################
+import os
+
+
+def test_model_weights_exist():
+    """Tests for existence of classifier model."""
+    assert os.path.isfile("best_model.h5")
+
+
+def test_heroku_setup():
+    """Tests for existance of setup file with personal email"""
+    assert os.path.isfile("setup.sh")

utils.py

@@ -0,0 +1,43 @@
+"""
+This module contains custom metric functions for binary classification models using TensorFlow and Keras.
+
+Functions:
+- f1(y_true, y_pred): Computes the F1 score metric for binary classification models.
+- recall(y_true, y_pred): Computes the recall metric for binary classification models.
+- precision(y_true, y_pred): Computes the precision metric for binary classification models.
+"""
+
+import tensorflow as tf
+from tensorflow.keras import backend as K
+
+
+def f1(y_true, y_pred):
+    y_pred = K.round(y_pred)
+    tp = K.sum(K.cast(y_true * y_pred, "float"), axis=0)
+    tn = K.sum(K.cast((1 - y_true) * (1 - y_pred), "float"), axis=0)
+    fp = K.sum(K.cast((1 - y_true) * y_pred, "float"), axis=0)
+    fn = K.sum(K.cast(y_true * (1 - y_pred), "float"), axis=0)
+
+    precision = tp / (tp + fp + K.epsilon())
+    recall = tp / (tp + fn + K.epsilon())
+    f1_score = 2 * precision * recall / (precision + recall + K.epsilon())
+
+    return f1_score
+
+
+def recall(y_true, y_pred):
+    y_pred = K.round(y_pred)
+    tp = K.sum(K.cast(y_true * y_pred, "float"), axis=0)
+    fn = K.sum(K.cast(y_true * (1 - y_pred), "float"), axis=0)
+
+    recall = tp / (tp + fn + K.epsilon())
+    return recall
+
+
+def precision(y_true, y_pred):
+    y_pred = K.round(y_pred)
+    tp = K.sum(K.cast(y_true * y_pred, "float"), axis=0)
+    fp = K.sum(K.cast((1 - y_true) * y_pred, "float"), axis=0)
+
+    precision = tp / (tp + fp + K.epsilon())
+    return precision