Skip to main content

​
Overview

TrashDetect.py is a command-line interface for real-time waste detection using your webcam. It provides a lightweight, OpenCV-based detection window with bounding boxes and confidence scores.
TrashDetect.py requires a connected camera (default: camera index 0) and the trained YOLOv8 model at Modelos/best.pt.

​
Running the CLI Detector

1

Navigate to Project Directory

Open your terminal and navigate to the Reciclaje AI source directory:
cd /path/to/reciclaje-ai/source
2

Run TrashDetect.py

Execute the detection script:
python TrashDetect.py
The script will immediately start capturing video from your default webcam (camera index 0).
3

Interact with Detection Window

A window titled “Waste Detect” will appear showing:
  • Live video feed from your camera
  • Red bounding boxes around detected waste items
  • Classification labels (Metal, Glass, Plastic, Carton, Medical)
  • Confidence percentages
To exit, press ESC (key code 27).

​
Detection Output

The CLI detector provides two types of output:

​
Visual Output

The OpenCV window displays:
  • Bounding boxes: Red rectangles (BGR: 0, 0, 255) around detected objects
  • Labels: Class name and confidence percentage above each detection
  • Font: HERSHEY_COMPLEX with size 1 and thickness 2

​
Console Output

Real-time detection information is printed to the console: 
Clase: 2 Confidence: 1
Clase: 0 Confidence: 1
Clase: 1 Confidence: 1
Where:
  • Clase (Class): Integer ID (0=Metal, 1=Glass, 2=Plastic, 3=Carton, 4=Medical)
  • Confidence: Detection confidence (0-1 scale, displayed as ceiling value)

​
Detected Waste Classes

The model can detect five types of waste materials:

Metal

Class ID: 0Recyclable metals like aluminum cans, steel containers

Glass

Class ID: 1Glass bottles, jars, and containers

Plastic

Class ID: 2Plastic bottles, containers, and packaging

Carton

Class ID: 3Cardboard boxes, paper cartons

Medical

Class ID: 4Medical waste requiring special disposal

​
Code Structure

Here’s the complete TrashDetect.py implementation: 
# Importamos librerias
from ultralytics import YOLO
import cv2
import math

# Modelo
model = YOLO('Modelos/best.pt')

# Cap
cap = cv2.VideoCapture(0)
cap.set(3, 1280)
cap.set(4, 720)

# Clases
clsName = ['Metal', 'Glass', 'Plastic', 'Carton', 'Medical']

# Inference
while True:
    # Frames
    ret, frame = cap.read()

    # Yolo | AntiSpoof
    results = model(frame, stream=True, verbose=False)
    for res in results:
        # Box
        boxes = res.boxes
        for box in boxes:
            # Bounding box
            x1, y1, x2, y2 = box.xyxy[0]
            x1, y1, x2, y2 = int(x1), int(y1), int(x2), int(y2)

            # Error < 0
            if x1 < 0: x1 = 0
            if y1 < 0: y1 = 0
            if x2 < 0: x2 = 0
            if y2 < 0: y2 = 0

            # Class
            cls = int(box.cls[0])

            # Confidence
            conf = math.ceil(box.conf[0])
            print(f"Clase: {cls} Confidence: {conf}")

            if conf > 0:
                # Draw
                cv2.rectangle(frame, (x1, y1), (x2, y2), (0, 0, 255), 2)
                cv2.putText(frame, f'{clsName[cls]} {int(conf * 100)}%', (x1, y1 - 20),
                            cv2.FONT_HERSHEY_COMPLEX, 1, (0, 0, 255), 2)

    # Show
    cv2.imshow("Waste Detect", frame)

    # Close
    t = cv2.waitKey(5)
    if t == 27:
        break

cap.release()
cv2.destroyAllWindows()

​
Key Features

​
Model Inference

TrashDetect.py:23
results = model(frame, stream=True, verbose=False)
  • stream=True: Enables generator-based inference for memory efficiency
  • verbose=False: Suppresses detailed logging output

​
Bounding Box Validation

TrashDetect.py:33-36
if x1 < 0: x1 = 0
if y1 < 0: y1 = 0
if x2 < 0: x2 = 0
if y2 < 0: y2 = 0
Ensures bounding box coordinates stay within frame boundaries.

​
Confidence Thresholding

TrashDetect.py:45
if conf > 0:
Only displays detections with positive confidence scores to reduce false positives.

​
Keyboard Controls

Press ESC at any time to exit the detection loop and close the application gracefully.
KeyAction
ESC (27)Exit application and release camera

​
Troubleshooting

If you see a camera initialization error:
  1. Ensure your webcam is connected
  2. Check if another application is using the camera
  3. Try changing the camera index in line 10:
cap = cv2.VideoCapture(1)  # Try index 1, 2, etc.
Verify the model path exists:
ls Modelos/best.pt
If missing, ensure you’ve trained the model or downloaded the pre-trained weights.
Detection accuracy depends on:
  • Proper lighting conditions
  • Camera positioning and angle
  • Distance to objects (optimal: 30-100cm)
  • Object visibility and size in frame
See Camera Configuration for optimization tips.
To improve performance:
  1. Reduce camera resolution (see Camera Configuration)
  2. Ensure GPU support is enabled for YOLOv8
  3. Close other resource-intensive applications

​
Next Steps

GUI Application

Learn about the full-featured Tkinter GUI interface

Camera Configuration

Optimize camera settings for best detection results

Build docs developers (and LLMs) love

Get started for freeTalk to us