Upload 13 files

Dockerfile

@@ -0,0 +1,11 @@
+FROM python:3.10
+
+WORKDIR /code
+
+COPY ./requirements.txt /code/requirements.txt
+
+RUN pip install --no-cache-dir --upgrade -r /code/requirements.txt
+
+COPY ./app /code/app
+
+CMD ["uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "8080"]

README.md

@@ -1,3 +1,54 @@
----
-license: mit
----
+# Saliency Inference API Template
+
+This is an API and Streamlit app to interact with a saliency model. The API is built using FastAPI and the Streamlit app is built using Streamlit. The API is built to be run in a Docker container.
+
+## Setup
+
+### Install dependencies
+
+```bash
+pip install -r requirements.txt
+```
+
+### Run the API
+
+```bash
+uvicorn main:app --reload --workers 1 --host 0.0.0.0 --port 8080
+```
+
+This will run the FastAPI server on port 8080.
+
+### (Alternative) Run the API in a Docker container
+
+
+```bash
+docker build -t ds-api-template .
+docker run -p 8080:8080 ds-api-template
+```
+
+You can test this is running by executing the same `curl` command as above, which should return the same response.
+
+NOTE: You will need to have Docker installed on your machine. To install Docker, follow the instructions [here](https://docs.docker.com/get-docker/).
+
+## Run the Streamlit App
+
+Once you've set up the API, you can run the Streamlit app to interact with the API.
+
+To run the Streamlit app, run the following command:
+
+```bash
+streamlit run app.py
+```
+
+You will need to have Streamlit installed on your machine. To install Streamlit, run the following command:
+
+```bash
+pip install streamlit
+```
+
+You will also need to update a `secrets.toml` file in a `.streamlit` directory at the root of the repo. This file should contain the following:
+
+```toml
+api_host = "http://localhost:8501"
+password = "<INSERT DESIRED PASSWORD HERE>"
+```

app/main.py

@@ -0,0 +1,34 @@
+from fastapi import FastAPI, File
+from fastapi.middleware.cors import CORSMiddleware
+
+from .model import predict
+import json
+
+app = FastAPI()
+
+# CORS
+origins = [
+    "http://localhost:8080",
+    "http://localhost"
+]
+
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=origins,
+    allow_credentials=True,
+    allow_methods=["POST"],
+    allow_headers=["*"],
+)
+
+@app.post("/predict")
+def img_object_detection_to_img(file: bytes = File(...)):
+    """
+    Object Detection from an image plot bbox on image
+
+    Args:
+        file (bytes): The image file in bytes format.
+    Returns:
+        The json representation of the prediction
+    """
+    prediction = predict(file)
+    return json.dumps(prediction.tolist())

app/model.py

@@ -0,0 +1,41 @@
+import tensorflow as tf
+from PIL import Image
+import io
+
+imported = tf.saved_model.load("./app")
+imported = imported.signatures["serving_default"]
+
+def get_image_from_bytes(binary_image: bytes) -> Image:
+    """Convert image from bytes to PIL RGB format
+    
+    Args:
+        binary_image (bytes): The binary representation of the image
+    
+    Returns:
+        PIL.Image: The image in PIL RGB format
+    """
+    input_image = Image.open(io.BytesIO(binary_image)).convert("RGB")
+    return input_image
+
+def predict(input_image):
+    """Reads file and returns prediction
+
+    Args:
+        x (_type_): _description_
+
+    Returns:
+        _type_: _description_
+    """
+    tensor = tf.io.decode_image(input_image, channels=3)
+
+    inference_shape = (240, 320)
+    original_shape = tensor.shape[:2]
+
+    input_tensor = tf.expand_dims(tensor, axis=0)
+
+    input_tensor = tf.image.resize(input_tensor, inference_shape,
+                               preserve_aspect_ratio=True)
+    saliency = imported(input_tensor)["output"]
+
+    saliency = tf.image.resize(saliency, original_shape)
+    return saliency.numpy()[0]

app/saved_model.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:646e0f343c4357e828f2569bef2f2bf288449fe68f7e4fb43e076f2e3b094e3d
+size 99858975

convert_model.py

@@ -0,0 +1,33 @@
+# use this script to convert any of the models saved to be
+# compatible with tf2: https://drive.google.com/drive/folders/1GI7i6GpfI-FoklP3vCc6vxe3T9nk3V2n
+
+import tensorflow as tf
+from tensorflow.python.saved_model import signature_constants, tag_constants
+
+export_dir = "./app/"
+# update the below line to point at the desired model downloaded
+# from the above google drive link
+graph_pb = "./app/model_salicon_cpu.pb"
+
+with tf.io.gfile.GFile(graph_pb, "rb") as f:
+    graph_def = tf.compat.v1.GraphDef()
+    graph_def.ParseFromString(f.read())
+
+sig = {}
+
+builder = tf.compat.v1.saved_model.Builder(export_dir)
+
+with tf.compat.v1.Session(graph=tf.Graph()) as sess:
+    tf.import_graph_def(graph_def, name="")
+    g = tf.compat.v1.get_default_graph()
+
+    input = g.get_tensor_by_name("input:0")
+    output = g.get_tensor_by_name("output:0")
+
+    sig_key = signature_constants.DEFAULT_SERVING_SIGNATURE_DEF_KEY
+    sig[sig_key] = tf.compat.v1.saved_model.predict_signature_def({"input": input},
+                                                                  {"output": output})
+    builder.add_meta_graph_and_variables(sess,
+                                         [tag_constants.SERVING],
+                                         signature_def_map=sig)
+builder.save()

main.py

@@ -0,0 +1,34 @@
+from fastapi import FastAPI, File
+from fastapi.middleware.cors import CORSMiddleware
+
+from .model import predict
+import json
+
+app = FastAPI()
+
+# CORS
+origins = [
+    "http://localhost:8080",
+    "http://localhost"
+]
+
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=origins,
+    allow_credentials=True,
+    allow_methods=["POST"],
+    allow_headers=["*"],
+)
+
+@app.post("/predict")
+def img_object_detection_to_img(file: bytes = File(...)):
+    """
+    Object Detection from an image plot bbox on image
+
+    Args:
+        file (bytes): The image file in bytes format.
+    Returns:
+        The json representation of the prediction
+    """
+    prediction = predict(file)
+    return json.dumps(prediction.tolist())

model.py

@@ -0,0 +1,41 @@
+import tensorflow as tf
+from PIL import Image
+import io
+
+imported = tf.saved_model.load("./app")
+imported = imported.signatures["serving_default"]
+
+def get_image_from_bytes(binary_image: bytes) -> Image:
+    """Convert image from bytes to PIL RGB format
+    
+    Args:
+        binary_image (bytes): The binary representation of the image
+    
+    Returns:
+        PIL.Image: The image in PIL RGB format
+    """
+    input_image = Image.open(io.BytesIO(binary_image)).convert("RGB")
+    return input_image
+
+def predict(input_image):
+    """Reads file and returns prediction
+
+    Args:
+        x (_type_): _description_
+
+    Returns:
+        _type_: _description_
+    """
+    tensor = tf.io.decode_image(input_image, channels=3)
+
+    inference_shape = (240, 320)
+    original_shape = tensor.shape[:2]
+
+    input_tensor = tf.expand_dims(tensor, axis=0)
+
+    input_tensor = tf.image.resize(input_tensor, inference_shape,
+                               preserve_aspect_ratio=True)
+    saliency = imported(input_tensor)["output"]
+
+    saliency = tf.image.resize(saliency, original_shape)
+    return saliency.numpy()[0]

requirements.txt

@@ -0,0 +1,7 @@
+fastapi==0.103.2
+uvicorn==0.23.2
+tensorflow
+python-multipart
+Pillow
+streamlit
+matplotlib

saved_model.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:646e0f343c4357e828f2569bef2f2bf288449fe68f7e4fb43e076f2e3b094e3d
+size 99858975

streamlit_viz.py

@@ -0,0 +1,71 @@
+"""App to visualize saliency maps for images.
+To run, use:
+streamlit run streamlit_viz.py
+"""
+
+import streamlit as st
+import pandas as pd
+import numpy as np
+import requests
+import hmac
+import json
+import matplotlib.pyplot as plt
+import matplotlib.image as mpimg
+
+from PIL import Image
+
+st.set_option('deprecation.showPyplotGlobalUse', False)
+
+def check_password():
+    """Returns `True` if the user had the correct password."""
+
+    def password_entered():
+        """Checks whether a password entered by the user is correct."""
+        if hmac.compare_digest(st.session_state["password"], st.secrets["password"]):
+            st.session_state["password_correct"] = True
+            del st.session_state["password"]  # Don't store the password.
+        else:
+            st.session_state["password_correct"] = False
+
+    # Return True if the passward is validated.
+    if st.session_state.get("password_correct", False):
+        return True
+
+    # Show input for password.
+    st.text_input(
+        "Password", type="password", on_change=password_entered, key="password"
+    )
+    if "password_correct" in st.session_state:
+        st.error("😕 Password incorrect")
+    return False
+
+
+if not check_password():
+    st.stop()  # Do not continue if check_password is not True.
+
+st.title("Saliency Map Visualizer")
+
+st.markdown(
+    """
+    This is a demo of the Saliency Map Visualizer. To use it, upload an image
+    and click the button below. Please note, it may take up to 20 seconds to visualise.
+    """
+)
+
+# get host from secrets
+api_host = st.secrets["api_host"]
+
+uploaded_file = st.file_uploader("Choose an image...", type=(["jpg", "jpeg", "png"]))
+
+if uploaded_file is not None:
+    file = {'file': uploaded_file.read()}
+    st.write("")
+    st.write("Classifying...")
+    response = requests.post(api_host, files=file)
+    arr = np.asarray(json.loads(response.json()))
+    st.write("Done!")
+    # Show plt plots
+    plt.imshow(Image.open(uploaded_file))
+    plt.imshow(arr, alpha=0.6)
+    plt.axis('off')
+    st.pyplot()