Refactor app.py to use gemini client for bird feather detection

app.py

@@ -1,42 +1,44 @@
 import gradio as gr
-import pandas as pd
-import numpy as np
-import os
-from PIL import Image
-import requests
-from io import BytesIO
-from imageai.Prediction import ImagePrediction
 import csv
+import gemini
+import base64
+import httpx
 
-# Load the protected-birds.csv file
-protected_birds = pd.read_csv('protected-birds.csv')
-protected_birds = protected_birds.dropna()
-protected_birds = protected_birds['Common Name'].str.lower().values
+# Load the protected-birds.txt file into a dictionary
+protected_birds_dict = {}
+with open('protected-birds.txt', 'r') as txtfile:
+    for line in txtfile:
+        bird, confidence = line.strip().split(',')
+        protected_birds_dict[bird.lower()] = float(confidence)
 
-# Load the image ai model
-model = ImagePrediction()
-model.setModelTypeAsResNet()
-model.setModelPath("resnet50_weights_tf_dim_ordering_tf_kernels.h5")
-model.loadModel()
+# Initialize the gemini client
+def init_client(api_key):
+    return gemini.Client(api_key)
 
 # Function to check if the feather is from a protected bird
-def check_feather(image):
+def check_feather(image_url, api_key):
     # Load the image
-    response = requests.get(image)
-    img = Image.open(BytesIO(response.content))
-    img.save('image.jpg')
+    image_media_type = "image/jpeg"
+    image_data = base64.b64encode(httpx.get(image_url).content).decode("utf-8")
     
     # Make a prediction
-    model_predictions, probabilities = model.predictImage('image.jpg', result_count=5)
+    client = init_client(api_key)
+    response = client.predict(
+        model_id="text-bison-001",
+        inputs={
+            "image": image_data,
+            "question": "Describe this image."
+        }
+    )
+    model_predictions = response["candidates"][0]["output"]
     
     # Check if the feather is from a protected bird
     protected = False
     protected_birds_list = []
-    for bird in protected_birds:
-        for prediction in model_predictions:
-            if bird in prediction.lower():
-                protected = True
-                protected_birds_list.append([bird, prediction, probabilities[model_predictions.index(prediction)]])
+    for bird, confidence in protected_birds_dict.items():
+        if bird in model_predictions.lower():
+            protected = True
+            protected_birds_list.append([bird, confidence, model_predictions])
     
     if protected:
         # Display the results
@@ -48,10 +50,11 @@ def check_feather(image):
             # Prompt user to make a report
             gr.Interface(fn=lambda: "If confident, report to: https://www.fws.gov/contact-us", inputs=user_verification, outputs="text").launch()
     else:
-        gr.Interface(fn=lambda: "No protected bird detected", inputs=image, outputs="text").launch()
+        gr.Interface(fn=lambda: "No protected bird detected", inputs=image_url, outputs="text").launch()
 
     return protected, protected_birds_list
 
 # Gradio interface
-image = gr.inputs.Image()
-gr.Interface(fn=check_feather, inputs=image, outputs="text").launch()
+image_url = gr.inputs.Textbox()
+api_key = gr.inputs.Textbox(label="Google API Key")
+gr.Interface(fn=check_feather, inputs=[image_url, api_key], outputs="text").launch()

package.txt

@@ -1,8 +1,3 @@
-gradio
-pandas
-numpy
-Pillow
-requests
-imageai
-tensorflow==2.4.0
-h5py==2.10.0
+transformers==4.26.0
+datasets==2.4.0
+sentencepiece==0.1.96

list.csv → protected-birds.txt

@@ -1,4 +1,3 @@
-ENGLISH_NAME
 Black-bellied Whistling-Duck
 West Indian Whistling-Duck
 Fulvous Whistling-Duck

requirements.txt

@@ -1,10 +1,6 @@
 gradio
-pandas
-numpy
-Pillow
-requests
-imageai==2.1.6
-tensorflow>=2.4.0
-pytest
-mock
-coverage
+anthropic
+httpx
+transformers==4.26.0
+datasets==2.4.0
+sentencepiece==0.1.96