Pre-trained detection models in Computer Vision

Introduction

Pre-trained detection models help find objects in images quickly without needing to train from scratch. They save time and work by using knowledge learned from many images.

You want to find cars or people in photos without building a model yourself.

You need to quickly test object detection on new images.

You want to build an app that recognizes common objects like dogs or chairs.

You have limited data but want good detection results.

You want to learn how object detection works using ready models.

Syntax

Computer Vision

model = load_pretrained_model('model_name')
predictions = model.detect(image)

load_pretrained_model loads a ready-made detection model.

detect runs the model on an image and returns detected objects with locations.

Examples

Loads a Faster R-CNN model and detects objects in the image.

Computer Vision

model = load_pretrained_model('fasterrcnn_resnet50_fpn')
predictions = model.detect(image)

Loads a lightweight SSD MobileNet model for faster detection on mobile devices.

Computer Vision

model = load_pretrained_model('ssd_mobilenet_v2')
predictions = model.detect(image)

Sample Model

This code loads a pre-trained Faster R-CNN model, processes an image, and prints detected object labels with confidence scores above 0.5.

Computer Vision

import torchvision
import torch
from PIL import Image
import torchvision.transforms as T

# Load a pre-trained Faster R-CNN model
model = torchvision.models.detection.fasterrcnn_resnet50_fpn(pretrained=True)
model.eval()

# Load and prepare an image
image = Image.open('test_image.jpg')
transform = T.Compose([T.ToTensor()])
image_tensor = transform(image)

# Run detection
with torch.no_grad():
    predictions = model([image_tensor])

# Print detected labels and scores
labels = predictions[0]['labels']
scores = predictions[0]['scores']
print('Detected objects:')
for label, score in zip(labels, scores):
    if score > 0.5:
        print(f'Label ID: {label.item()}, Score: {score.item():.2f}')

Important Notes

Pre-trained models come with labels as numbers; you can map them to names using the model's label map.

Detection models return bounding boxes, labels, and confidence scores for each object.

Using a threshold on scores helps ignore weak detections.

Summary

Pre-trained detection models let you find objects in images without training.

They save time and work by using knowledge from large datasets.

You can use them to quickly build apps or learn object detection basics.