Addded files for tracking different fields using multiple GPUs

2_detect_red_all.py

@@ -0,0 +1,101 @@
+import cv2
+import os
+import pandas as pd
+import numpy as np
+from os.path import splitext, basename, join
+from datetime import datetime, timedelta
+from glob import glob
+
+# Function to detect red light in the ROI
+def detect_red_light(image):
+    lower_red = np.array([0, 120, 70])
+    upper_red = np.array([10, 255, 255])
+    hsv = cv2.cvtColor(image, cv2.COLOR_BGR2HSV)
+    mask = cv2.inRange(hsv, lower_red, upper_red)
+    return cv2.countNonZero(mask) > 0
+
+# Function to manually draw the ROI
+def draw_roi(event, x, y, flags, param):
+    global traffic_light_box, drawing, top_left_pt, bottom_right_pt, roi_selected
+    if event == cv2.EVENT_LBUTTONDOWN:
+        drawing = True
+        top_left_pt = (x, y)
+    elif event == cv2.EVENT_LBUTTONUP:
+        drawing = False
+        bottom_right_pt = (x, y)
+        roi_selected = True
+
+# Paths
+videos_path = 'C:/Users/marya1/Box/MnDOT DNRTOR Project/Meenakshi/videos'
+output_path = 'C:/Users/marya1/Box/MnDOT DNRTOR Project/Meenakshi/detect_red'
+#videos_path = 'C:/Users/ASUS/Box/MnDOT DNRTOR Project/Pratik'
+#output_path = 'C:/Users/ASUS/Box/MnDOT DNRTOR Project/Pratik/output'
+#videos_path = 'E:/MnDOT/Videos'
+#output_path = 'E:/MnDOT/Videos/output'
+
+
+# List all video files
+video_files = glob(join(videos_path, '*.mp4'))
+
+for video_path in video_files:
+    # Initialize variables for each video
+    traffic_light_box = None
+    roi_selected = False
+
+    # Display video and select ROI
+    cap = cv2.VideoCapture(video_path)
+    ret, first_frame = cap.read()
+    if not ret:
+        print("Failed to read video:", video_path)
+        cap.release()
+        continue  # Skip to the next video
+
+    # Input for date and time
+    video_date = input(f"Enter the video start date (yyyy-mm-dd) for {basename(video_path)}: ")
+    video_start_time = input(f"Enter the video start time (hh-mm-ss) for {basename(video_path)}: ")
+    start_datetime = datetime.strptime(f"{video_date} {video_start_time}", "%Y-%m-%d %H-%M-%S")
+
+    cv2.namedWindow('Frame')
+    cv2.setMouseCallback('Frame', draw_roi)
+
+    print("Draw a bounding box around the traffic light and press 'Enter'.")
+    while True:
+        frame_copy = first_frame.copy()
+        if roi_selected:
+            cv2.rectangle(frame_copy, top_left_pt, bottom_right_pt, (0, 255, 0), 2)
+        cv2.imshow('Frame', frame_copy)
+        key = cv2.waitKey(1)
+        if key == 13:  # Enter key
+            traffic_light_box = (top_left_pt[0], top_left_pt[1], bottom_right_pt[0] - top_left_pt[0], bottom_right_pt[1] - top_left_pt[1])
+            break
+
+    cv2.destroyAllWindows()
+
+    df = pd.DataFrame(columns=['Frame_Number', 'Date', 'Time', 'Red_Light'])
+    frame_counter = 0
+    fps = cap.get(cv2.CAP_PROP_FPS)
+
+    while cap.isOpened():
+        ret, frame = cap.read()
+        if not ret:
+            break
+        current_time = start_datetime + timedelta(seconds=frame_counter / fps)
+        date_str = current_time.strftime("%Y-%m-%d")
+        time_str = current_time.strftime("%H:%M:%S")
+
+        traffic_light_roi = frame[traffic_light_box[1]:traffic_light_box[1]+traffic_light_box[3], 
+                                  traffic_light_box[0]:traffic_light_box[0]+traffic_light_box[2]]
+        red_light_detected = detect_red_light(traffic_light_roi)
+
+        new_row = {'Frame_Number': frame_counter, 'Date': date_str, 'Time': time_str, 'Red_Light': red_light_detected}
+        df = pd.concat([df, pd.DataFrame([new_row])], ignore_index=True)
+
+        frame_counter += 1
+
+    cap.release()
+
+    # Save the DataFrame to a CSV file
+    csv_file_name = splitext(basename(video_path))[0] + '.csv'
+    csv_path = join(output_path, csv_file_name)
+    df.to_csv(csv_path, index=False)
+    print(f"CSV saved at {csv_path}")

3_merge_detection_red.py

@@ -0,0 +1,55 @@
+import pandas as pd
+import os
+from glob import glob
+
+def merge_csv_with_timestamps(output_csv_dir, timestamps_csv_dir, merged_csv_dir):
+    # List all CSV files in the output CSV directory
+    output_csv_files = glob(os.path.join(output_csv_dir, '*.csv'))
+
+    for output_csv_path in output_csv_files:
+        # Derive the base file name without extension to match with timestamp file
+        base_file_name = os.path.basename(output_csv_path)
+
+        # Construct the expected timestamp CSV file path
+        timestamps_csv_path = os.path.join(timestamps_csv_dir, base_file_name)
+
+        # Check if the corresponding timestamps CSV exists
+        if not os.path.exists(timestamps_csv_path):
+            print(f"No matching timestamp file found for {base_file_name}. Skipping...")
+            continue
+
+        # Load the CSV files
+        output_df = pd.read_csv(output_csv_path)
+        timestamps_df = pd.read_csv(timestamps_csv_path)
+
+        # Merge the two dataframes on the Frame and Frame_Number columns
+        merged_df = pd.merge(output_df, timestamps_df, left_on='Frame', right_on='Frame_Number', how='left')
+
+        # Including all relevant columns and renaming
+        #final_df = merged_df[['Datetime', 'Frame', 'Track', 'Class', 'BBox', 'Time', 'Date', 'Red_Light']].copy()
+        final_df = merged_df[['Datetime', 'Frame', 'Track', 'Class', 'Class_ID','xmin', 'ymin','xmax','ymax', 'Time', 'Date', 'Red_Light']].copy()
+        final_df.rename(columns={'Frame': 'FrameNumber', 'Red_Light': 'Color'}, inplace=True)
+
+        # Handle 'Color' column
+        final_df['Color'].fillna('Not Red', inplace=True)
+        final_df['Color'] = final_df['Color'].apply(lambda x: 'Red' if x == True else 'Not Red')
+
+        # Construct path for saving the merged CSV
+        merged_csv_path = os.path.join(merged_csv_dir, base_file_name)
+
+        # Save the merged dataframe
+        final_df.to_csv(merged_csv_path, index=False)
+        print(f"Merged CSV saved to {merged_csv_path}")
+
+# Directories containing the CSV files
+#output_csv_dir = 'C:/Users/marya1/Box/MnDOT DNRTOR Project/Meenakshi/detect_objects'
+#timestamps_csv_dir = 'C:/Users/marya1/Box/MnDOT DNRTOR Project/Meenakshi/detect_red'
+#merged_csv_dir = 'C:/Users/marya1/Box/MnDOT DNRTOR Project/Meenakshi/final_csv'
+
+
+output_csv_dir = '/home/marya1/Documents/MnDoTNRToR/inference/detect_objects/processing'
+timestamps_csv_dir = '/home/marya1/Documents/MnDoTNRToR/inference/detect_red'
+merged_csv_dir = '/home/marya1/Documents/MnDoTNRToR/inference/final_csv'
+
+# Call the function with the directories
+merge_csv_with_timestamps(output_csv_dir, timestamps_csv_dir, merged_csv_dir)

3_merge_detection_red2.py

@@ -0,0 +1,39 @@
+import pandas as pd
+import os
+from glob import glob
+
+def merge_csv_with_timestamps(output_csv_dir, timestamps_csv_dir, merged_csv_dir):
+    output_csv_files = glob(os.path.join(output_csv_dir, '*.csv'))
+
+    for output_csv_path in output_csv_files:
+        base_file_name = os.path.basename(output_csv_path)
+        timestamps_csv_path = os.path.join(timestamps_csv_dir, base_file_name)
+
+        if not os.path.exists(timestamps_csv_path):
+            print(f"No matching timestamp file found for {base_file_name}. Skipping...")
+            continue
+
+        try:
+            output_df = pd.read_csv(output_csv_path, on_bad_lines='warn')
+            timestamps_df = pd.read_csv(timestamps_csv_path, on_bad_lines='warn')
+        except Exception as e:
+            print(f"Error reading {base_file_name}: {e}")
+            continue
+
+        merged_df = pd.merge(output_df, timestamps_df, left_on='Frame', right_on='Frame_Number', how='left')
+        final_df = merged_df[['Datetime', 'Frame', 'Track', 'Class', 'Class_ID', 'xmin', 'ymin', 'xmax', 'ymax', 'Time', 'Date', 'Red_Light']].copy()
+        final_df.rename(columns={'Frame': 'FrameNumber', 'Red_Light': 'Color'}, inplace=True)
+        final_df['Color'].fillna('Not Red', inplace=True)
+        final_df['Color'] = final_df['Color'].apply(lambda x: 'Red' if x == True else 'Not Red')
+
+        merged_csv_path = os.path.join(merged_csv_dir, base_file_name)
+        final_df.to_csv(merged_csv_path, index=False)
+        print(f"Merged CSV saved to {merged_csv_path}")
+
+
+output_csv_dir = '/home/marya1/Documents/MnDoTNRToR/inference/detect_objects/processing'
+timestamps_csv_dir = '/home/marya1/Documents/MnDoTNRToR/inference/detect_red'
+merged_csv_dir = '/home/marya1/Documents/MnDoTNRToR/inference/final_csv'
+
+# Use the directories as defined in the script
+merge_csv_with_timestamps(output_csv_dir, timestamps_csv_dir, merged_csv_dir)

4_detect_violations_confirmation.py

@@ -0,0 +1,126 @@
+import cv2
+import pandas as pd
+from os.path import join, basename, splitext
+from glob import glob
+
+# Define global variables for ROI drawing
+drawing = False
+roi_selected = False
+top_left_pt, bottom_right_pt = None, None
+
+def draw_roi(event, x, y, flags, param):
+    global drawing, roi_selected, top_left_pt, bottom_right_pt
+    if event == cv2.EVENT_LBUTTONDOWN:
+        drawing = True
+        top_left_pt = (x, y)
+    elif event == cv2.EVENT_LBUTTONUP:
+        drawing = False
+        bottom_right_pt = (x, y)
+        roi_selected = True
+
+def is_within_roi(xmin, ymin, xmax, ymax, roi):
+    roi_xmin, roi_ymin, roi_xmax, roi_ymax = roi
+    return xmin >= roi_xmin and xmax <= roi_xmax and ymin >= roi_ymin and ymax <= roi_ymax
+
+
+def count_vehicles_and_violations(csv_file, roi, valid_classes=[1, 2, 3, 5, 7], min_frames=60):
+    df = pd.read_csv(csv_file)
+    # Assuming 'Date' column is in 'YYYY-MM-DD' format and 'Time' in 'HH:MM:SS' format
+    df['DateTime'] = pd.to_datetime(df['Date'] + ' ' + df['Time'], errors='coerce')
+
+    red_light_phases = df[df['Color'] == 'Red'].groupby((df['Color'] != df['Color'].shift()).cumsum())
+    results = []
+
+    for _, phase in red_light_phases:
+        start_of_red_light = phase['DateTime'].iloc[0]
+        next_phase_start = phase['DateTime'].iloc[-1] + pd.Timedelta(seconds=1)
+
+        # Skip intervals less than 1 second
+        if (next_phase_start - start_of_red_light).total_seconds() <= 1:
+            continue
+
+        tracks_in_roi = df[(df['DateTime'] >= start_of_red_light) & (df['DateTime'] < next_phase_start)]
+        tracks_in_roi = tracks_in_roi[tracks_in_roi.apply(lambda row: is_within_roi(row['xmin'], row['ymin'], row['xmax'], row['ymax'], roi), axis=1)]
+        valid_tracks_in_roi = tracks_in_roi[tracks_in_roi['Class_ID'].isin(valid_classes)]
+
+        # Count frames for each track and filter out those with less than min_frames
+        frame_counts = valid_tracks_in_roi.groupby('Track').size()
+        tracks_more_than_min_frames = frame_counts[frame_counts >= min_frames].index
+        valid_tracks_in_roi = valid_tracks_in_roi[valid_tracks_in_roi['Track'].isin(tracks_more_than_min_frames)]
+
+        exiting_tracks = valid_tracks_in_roi.groupby('Track').last()
+        exiting_tracks = exiting_tracks[exiting_tracks.index.isin(tracks_more_than_min_frames)]
+        violations = exiting_tracks[exiting_tracks['Color'] == 'Red'].index.nunique()
+        violating_track_ids = exiting_tracks[exiting_tracks['Color'] == 'Red'].index.unique().tolist()
+        total_vehicles = valid_tracks_in_roi['Track'].nunique()
+
+        total_vehicles_complying = valid_tracks_in_roi['Track'].nunique() - violations
+        complying_track_ids = valid_tracks_in_roi[~valid_tracks_in_roi['Track'].isin(violating_track_ids)]['Track'].unique().tolist()
+
+        results.append({
+            'Start_of_Red_Light': start_of_red_light.strftime('%Y-%m-%d %H:%M:%S'),
+            'Beginning_of_Not_Red': next_phase_start.strftime('%Y-%m-%d %H:%M:%S'),
+            'Total_Vehicles': total_vehicles,
+            'Vehicles_Complying': total_vehicles_complying,
+            'Vehicles_Violating': violations,
+            'Complying Tracks': complying_track_ids,
+            'Violating Tracks': violating_track_ids,
+            'Compliance Rate': (total_vehicles - violations) / total_vehicles if total_vehicles > 0 else 1
+        })
+
+    results_df = pd.DataFrame(results)
+
+
+    if not results_df.empty:
+        overall_compliance_rate = (results_df['Total_Vehicles'].sum() - results_df['Vehicles_Violating'].sum()) / results_df['Total_Vehicles'].sum()
+        overall_summary = pd.DataFrame([{'Start_of_Red_Light': 'Overall', 'Beginning_of_Not_Red': '', 'Total_Vehicles': results_df['Total_Vehicles'].sum(), 'Vehicles_Violating': results_df['Vehicles_Violating'].sum(), 'Compliance Rate': overall_compliance_rate}])
+        results_df = pd.concat([results_df, overall_summary], ignore_index=True)
+
+    return results_df
+
+csv_path = '/home/marya1/Documents/MnDoTNRToR/inference/final_csv'
+output_path='/home/marya1/Documents/MnDoTNRToR/inference/detect_violations_trial'
+videos_path = '/home/marya1/Documents/MnDoTNRToR/videos/processing'
+
+video_files = glob(join(videos_path, '*.mp4'))
+
+for video_path in video_files:
+    cap = cv2.VideoCapture(video_path)
+    ret, first_frame = cap.read()
+    if not ret:
+        print(f"Failed to read video: {video_path}")
+        continue
+
+    # Reset RoI selection variables
+    drawing = False
+    roi_selected = False
+    top_left_pt, bottom_right_pt = None, None
+
+    cv2.namedWindow('Frame')
+    cv2.setMouseCallback('Frame', draw_roi)
+    print("Draw a bounding box around the area of interest and press 'Enter'.")
+
+    while True:
+        frame_copy = first_frame.copy()
+        if roi_selected:
+            cv2.rectangle(frame_copy, top_left_pt, bottom_right_pt, (0, 255, 0), 2)
+        cv2.imshow('Frame', frame_copy)
+        key = cv2.waitKey(1) & 0xFF
+        if key == 13:  # Enter key is pressed
+            break
+
+    if not roi_selected:
+        print(f"RoI not selected for video: {video_path}. Skipping...")
+        cv2.destroyAllWindows()
+        continue
+
+    roi = (top_left_pt[0], top_left_pt[1], bottom_right_pt[0], bottom_right_pt[1])
+    cv2.destroyAllWindows()
+    cap.release()
+
+    # Proceed with analysis using selected RoI
+    csv_file = join(csv_path, splitext(basename(video_path))[0] + '.csv')
+    analysis_results = count_vehicles_and_violations(csv_file, roi)
+    output_file = join(output_path, splitext(basename(video_path))[0] + '_violations.csv')
+    analysis_results.to_csv(output_file, index=False)
+    print(f"Analysis saved to {output_file}")