Object-detection/app.py

import os
import cv2
import json
from flask import Flask, request, jsonify, render_template, send_from_directory
from flask_cors import CORS
from werkzeug.utils import secure_filename
from ultralytics import YOLO

app = Flask(__name__)
CORS(app)

# Configuration
UPLOAD_FOLDER = 'uploads'
RESULT_FOLDER = 'static/results'
ALLOWED_EXTENSIONS = {'png', 'jpg', 'jpeg', 'gif', 'mp4', 'avi', 'mov'}

app.config['UPLOAD_FOLDER'] = UPLOAD_FOLDER
app.config['RESULT_FOLDER'] = RESULT_FOLDER

# Create directories if they don't exist
os.makedirs(UPLOAD_FOLDER, exist_ok=True)
os.makedirs(RESULT_FOLDER, exist_ok=True)

# Load YOLOv8 model (pre-trained on COCO)
# We use the nano version for speed in this environment
model = YOLO('yolov8n.pt')

def allowed_file(filename):
    return '.' in filename and filename.rsplit('.', 1)[1].lower() in ALLOWED_EXTENSIONS

@app.route('/')
def index():
    return render_template('index.html')

@app.route('/detect', methods=['POST'])
def detect_objects():
    if 'file' not in request.files:
        return jsonify({'error': 'No file part'}), 400
    
    file = request.files['file']
    if file.filename == '':
        return jsonify({'error': 'No selected file'}), 400
    
    if file and allowed_file(file.filename):
        filename = secure_filename(file.filename)
        filepath = os.path.join(app.config['UPLOAD_FOLDER'], filename)
        file.save(filepath)
        
        # Determine file type
        ext = filename.rsplit('.', 1)[1].lower()
        
        if ext in {'mp4', 'avi', 'mov'}:
            return process_video(filepath, filename)
        else:
            return process_image(filepath, filename)
    
    return jsonify({'error': 'File type not allowed'}), 400

def process_image(filepath, filename):
    # Run inference
    results = model(filepath)
    
    # Get the first result (since we only processed one image)
    result = results[0]
    
    # Save the plotted image (with bounding boxes)
    res_filename = 'res_' + filename
    res_path = os.path.join(app.config['RESULT_FOLDER'], res_filename)
    result.save(filename=res_path)
    
    # Count objects
    detections = []
    counts = {}
    
    for box in result.boxes:
        cls_id = int(box.cls[0])
        label = model.names[cls_id]
        conf = float(box.conf[0])
        detections.append({
            'label': label,
            'confidence': conf,
            'box': box.xyxy[0].tolist()
        })
        counts[label] = counts.get(label, 0) + 1
        
    return jsonify({
        'success': True,
        'type': 'image',
        'result_url': f'/static/results/{res_filename}',
        'objects': detections,
        'counts': counts,
        'total_count': len(detections)
    })

def process_video(filepath, filename):
    # For video, we'll run inference and save the output video
    # Note: Processing a full video can be slow. 
    # We'll use ultralytics' built-in video saving capability for simplicity
    
    res_filename = 'res_' + filename.split('.')[0] + '.mp4'
    res_path = os.path.join(app.config['RESULT_FOLDER'], res_filename)
    
    # Run inference on the video
    # Use project/name to control output location
    results = model(filepath, stream=True)
    
    # We need to collect overall counts for the video
    all_seen_objects = {}
    
    # To keep it simple and faster for the demo, we process every 5th frame if it's long?
    # No, let's just use the built-in save for now.
    
    # Run model.predict with save=True
    save_results = model.predict(filepath, save=True, project=app.config['RESULT_FOLDER'], name='vid_temp', exist_ok=True)
    
    # Find the saved video. Ultralytics saves it in a subfolder.
    # We want to move it to our RESULT_FOLDER with a predictable name.
    # Typically it goes to RESULT_FOLDER/vid_temp/filename
    raw_saved_path = os.path.join(app.config['RESULT_FOLDER'], 'vid_temp', filename)
    
    if os.path.exists(raw_saved_path):
        import shutil
        shutil.move(raw_saved_path, res_path)
        # Cleanup temp dir
        shutil.rmtree(os.path.join(app.config['RESULT_FOLDER'], 'vid_temp'))
    else:
        # Fallback if names differ (sometimes .avi becomes .mp4 etc)
        # Just check the folder
        temp_dir = os.path.join(app.config['RESULT_FOLDER'], 'vid_temp')
        if os.path.exists(temp_dir):
            files = os.listdir(temp_dir)
            if files:
                shutil.move(os.path.join(temp_dir, files[0]), res_path)
                shutil.rmtree(temp_dir)

    # For video summary, we'll just run a quick pass to get unique objects
    # (In a real app, you'd aggregate frame-by-frame)
    # Just return success for now with the URL
    
    return jsonify({
        'success': True,
        'type': 'video',
        'result_url': f'/static/results/{res_filename}',
        'message': 'Video processed successfully'
    })

if __name__ == '__main__':
    app.run(debug=True, port=5000)