# Import required modules import streamlit as st from ultralytics import YOLO from PIL import Image import os import json import logging import tempfile import pandas as pd import matplotlib.pyplot as plt st.set_page_config( page_title="Fish Detector", page_icon="🐟", layout="wide" ) sample_images_folder = "./images/sample_images" logging.basicConfig(level=logging.INFO) # Model loading model_folder = "./models" st.sidebar.title("🐟 Fish or No Fish Detector") st.sidebar.markdown(""" ### For more information: - Contact: Michael.Akridge@NOAA.gov - Visit the [GitHub repository](https://github.com/MichaelAkridge-NOAA/Fish-or-No-Fish-Detector/) """) # Display model links st.sidebar.markdown("### Model Links") st.sidebar.markdown("- [YOLO11 Fish Detector - Grayscale](https://huggingface.co/akridge/yolo11-fish-detector-grayscale)") st.sidebar.markdown("- [YOLO11 Segment Fish - Grayscale](https://huggingface.co/akridge/yolo11-segment-fish-grayscale)") model_name = st.sidebar.selectbox("Select a YOLO model", os.listdir(model_folder)) model_path = os.path.join(model_folder, model_name) if not os.path.exists(model_path): st.error(f"Model file not found at {model_path}. Please check your setup.") st.stop() model = YOLO(model_path) # Sidebar configuration st.sidebar.header("Model Parameters") confidence = st.sidebar.slider("Detection Confidence Threshold", 0.0, 1.0, 0.35) final_confidence = st.sidebar.slider("Final Yes/No Confidence Threshold", 0.0, 1.0, 0.5) # Title and description st.title("🐟 Fish or No Fish Detector") st.write(""" Is there a fish 🐟 or not? Upload one or more images to detect fish. Using a trained [Ultralytics YOLO11 Model](https://github.com/ultralytics/ultralytics) for its object detection capabilities. """) # Custom CSS for button and uploader alignment st.markdown(""" """, unsafe_allow_html=True) # Custom CSS for default button styling st.markdown(""" """, unsafe_allow_html=True) # Load sample images function def load_sample_images(): return [os.path.join(sample_images_folder, img) for img in os.listdir(sample_images_folder) if img.lower().endswith(('png', 'jpg', 'jpeg'))] # Prediction function def run(image_path): results = model.predict(image_path, conf=confidence) boxes = [] fish_count = 0 confidences = [] for box in results[0].boxes: x1, y1, x2, y2 = box.xyxy[0].tolist() conf = box.conf[0].item() class_id = int(box.cls[0].item()) class_label = model.names[class_id].lower() # Normalize to lowercase if class_label == "fish" and conf > confidence: fish_count += 1 confidences.append(conf) boxes.append({"x1": x1, "y1": y1, "x2": x2, "y2": y2, "confidence": conf, "class_id": class_id, "class_label": class_label}) return results[0].plot()[:, :, ::-1], {"fish_count": fish_count, "confidences": confidences} # Process images function with directory creation # Reusable function to handle multiple image uploads and display results def process_images(uploaded_files): all_detections = [] result_images = [] summary_data = [] confidences = [] temp_dir = tempfile.gettempdir() for uploaded_file in uploaded_files: if isinstance(uploaded_file, str): # Check if it's a sample image path image_path = uploaded_file image = Image.open(image_path) else: image = Image.open(uploaded_file) image_path = os.path.join(temp_dir, f"{uploaded_file.name}") image.save(image_path) st.write(f"Detecting in {os.path.basename(image_path)}...") with st.spinner('Running detection...'): result_image, detection_metadata = run(image_path) if result_image is not None: result_images.append((result_image, os.path.basename(image_path))) all_detections.append(detection_metadata) summary_data.append({ "image_name": os.path.basename(image_path), "fish_detected": detection_metadata["fish_count"] > 0, "fish_count": detection_metadata["fish_count"] }) confidences.extend(detection_metadata["confidences"]) # Display fish status fish_detected = detection_metadata['fish_count'] > 0 fish_status = f"YES 🐟" if fish_detected else f"NO" st.markdown(f"**Summary for {os.path.basename(image_path)}:** Fish detected: {fish_status}", unsafe_allow_html=True) # Display images side by side col1, col2 = st.columns(2) with col1: st.image(image, caption=f"Uploaded Image - {os.path.basename(image_path)}", use_column_width=True) with col2: st.image(result_image, caption=f"Detection Results - {os.path.basename(image_path)}", use_column_width=True) st.success(f"Detection completed for {os.path.basename(image_path)} successfully! 🐟") else: st.warning(f"No marine ecosystems detected in {os.path.basename(image_path)}.") st.session_state["all_detections"] = all_detections return summary_data, confidences # Function to display a summary table and scatter plot side by side with image labels def display_summary(summary_data, confidences): if summary_data: df = pd.DataFrame(summary_data) col1, col2 = st.columns(2) with col1: st.subheader("Summary of Detections") st.table(df[["image_name", "fish_count"]]) with col2: st.subheader("Fish Detection Confidence Levels") fig, ax = plt.subplots() confidence_index = 0 for i, row in df.iterrows(): num_confidences_for_image = len([c for c in confidences[confidence_index:confidence_index + row["fish_count"]]]) for j in range(num_confidences_for_image): if confidence_index < len(confidences): ax.scatter(confidence_index, confidences[confidence_index], c='blue') ax.text(confidence_index, confidences[confidence_index], row['image_name'], fontsize=10, ha='center', va='bottom', rotation=0) confidence_index += 1 ax.axhline(final_confidence, color='red', linestyle='--', label=f'Final Threshold ({final_confidence})') ax.set_xlabel('Detections') ax.set_ylabel('Confidence Level') ax.legend(loc='lower left') st.pyplot(fig) if st.session_state.get("all_detections"): json_data = json.dumps(st.session_state["all_detections"], indent=4) st.download_button( label="Download Results as JSON & Reset", data=json_data, file_name="all_detections.json", mime="application/json", key="download_json_bottom" ) # Image uploader with multiple file support st.markdown('