camera impl

app.py

@@ -1,27 +1,70 @@
 import streamlit as st
 from ultralytics import YOLO
 import tempfile
+import pandas as pd
 
 # Load the model
 model = YOLO('best.pt')
 
-st.title('YOLO Object Detection')
+st.title('Spare-it Recyclable Detection Model')
 
-uploaded_file = st.file_uploader("Upload an image", type=['jpg', 'jpeg', 'png'])
+st.write("""
+This detection model is designed to identify recyclable items within images, providing high detection capabilities across 20 categories. Below are the performance metrics for specific recyclable materials.
+""")
 
-if uploaded_file is not None:
+# Hardcoded table with performance metrics
+data = {
+    'Category': [
+        'Overall', 'Shredded Paper', 'Metal Can', 'Aluminum Foil', 'Paper Cup',
+        'Office Paper', 'Paper Towel/Napkins/Tissue', 'Glass Bottles', 'Plastic Drink Bottle',
+        'Compostable Fiber Ware', 'Fruits And Veggies', 'Clear Plastic Cup', 'Black Plastic',
+        'Cardboard Coffee Cup Sleeve', 'Snack or Candy Bag or Wrapper', 'Plastic Lid except black',
+        'Filled Bag', 'Empty Plastic Bag', 'Clean Paper Plate', 'Shelf Stable Carton',
+        'Clean Cardboard', 'Empty Paper Bag'
+    ],
+    'Total Images': [1667] * 22,
+    'Samples': [12871, 16, 1227, 58, 405, 1588, 4304, 58, 378, 244, 635, 245, 136, 74, 1018, 509, 287, 529, 43, 134, 587, 396],
+    'Precision': [0.527, 0.845, 0.79, 0.674, 0.607, 0.564, 0.562, 0.562, 0.559, 0.52, 0.518, 0.513, 0.469, 0.467, 0.462, 0.459, 0.428, 0.422, 0.416, 0.414, 0.41, 0.402],
+    'Recall': [0.451, 0.5, 0.801, 0.414, 0.568, 0.409, 0.518, 0.5, 0.574, 0.623, 0.378, 0.551, 0.485, 0.343, 0.333, 0.344, 0.37, 0.308, 0.256, 0.396, 0.419, 0.371],
+    'mAP50': [0.448, 0.5, 0.855, 0.434, 0.595, 0.437, 0.533, 0.511, 0.578, 0.593, 0.377, 0.465, 0.479, 0.363, 0.339, 0.321, 0.37, 0.279, 0.275, 0.433, 0.35, 0.328],
+    'mAP50-95': [0.315, 0.429, 0.667, 0.307, 0.44, 0.272, 0.315, 0.298, 0.386, 0.489, 0.23, 0.331, 0.367, 0.244, 0.211, 0.232, 0.248, 0.158, 0.232, 0.296, 0.256, 0.215]
+}
+df = pd.DataFrame(data)
+
+st.dataframe(df, hide_index=True)
+
+input_method = st.radio("Choose the input method:", ("Upload an Image", "Take a Picture"))
+
+if input_method == "Upload an Image":
+    image_data = st.file_uploader("Upload an image", type=['jpg', 'jpeg', 'png'])
+
+elif input_method == "Take a Picture":
+    image_data = st.camera_input("Take a picture")
+
+if image_data is not None:
+    # make temp image file to run detection
     with tempfile.NamedTemporaryFile(delete=False, suffix='.jpg') as tmp_file:
-        tmp_file.write(uploaded_file.getvalue())
-        uploaded_image_path = tmp_file.name
+        tmp_file.write(image_data.getvalue())
+        image_path = tmp_file.name
+
+    results = model(image_path)
 
-    # Display the uploaded image
-    st.image(uploaded_image_path, caption='Uploaded Image', use_column_width=True)
+    category_names = results[0].names
+    predictions = {}
+    for box in results[0].boxes:
+        cls_id = int(box.cls)
+        conf = float(box.conf)
+        if cls_id in category_names:
+            if category_names[cls_id] in predictions:
+                predictions[category_names[cls_id]].append(conf)
+            else:
+                predictions[category_names[cls_id]] = [conf]
 
-    # Perform inference and save the result
-    results = model(uploaded_image_path)
-    saved_image_path = 'predicted_image.jpg'  # Define the path where the image will be saved
-    for result in results:
-        result.save(filename=saved_image_path)  # Save the predicted image
+    st.write(f"Total {len(results[0].boxes.cls)} objects found.")
+    for category, confidences in predictions.items():
+        st.write(f"{len(confidences)} {category}: {['{:.2f}'.format(c) for c in confidences]}")
 
-    # Display the image with predictions
-    st.image(saved_image_path, caption='Predicted Image', use_column_width=True)
+    with tempfile.NamedTemporaryFile(delete=False, suffix='.jpg') as output_tmp:
+        for result in results:
+            result.save(filename=output_tmp.name)
+        st.image(output_tmp.name, caption='Predicted Image', use_column_width=True)