Install YOLOv5

YOLO performs detection in a single forward pass—making it well-suited for real-time applications. To understand it effectively, visualize bounding boxes, class labels, confidence scores, and Non-Maximum Suppression (NMS) together on the same diagram. This article first establishes the big picture: what problem it solves, what its core components are, and which types of tasks it fits best.

I begin by fixing the input image size, then tune the confidence threshold and NMS threshold. Without recording these thresholds, detection results become difficult—or impossible—to reproduce.

In the previous article on DenseNet application examples, we explored DenseNet’s superior performance in image classification tasks and demonstrated how to train and run inference with this model in practice. In this article, we shift focus to the YOLO (You Only Look Once) model’s application in segmentation tasks—specifically, how YOLO enables the integration of real-time object detection and image segmentation.

Overview of YOLO

YOLO is an efficient, real-time object detection model. Its defining characteristic is framing object detection as a regression problem: a single neural network directly predicts bounding boxes and class probabilities in one pass. This design allows YOLO to achieve high accuracy while maintaining computational efficiency.

While reading this article, treat the progression “YOLO Overview → How YOLO Works → YOLO for Segmentation → YOLOv5 and Segmentation” as a verification checklist: first clarify the topic, logical flow, and validation points; only then revisit concrete examples, code snippets, or evaluation metrics for cross-checking.

How YOLO Works

YOLO divides the input image into an $S \times S$ grid. Each grid cell is responsible for predicting objects whose center falls within that cell. For each cell, the model generates multiple candidate bounding boxes along with associated confidence scores.

1. Architectural Framework: YOLO’s network architecture is typically built upon Convolutional Neural Networks (CNNs), with the final layer outputting predicted bounding boxes and class probabilities.
2. Loss Function: YOLO’s loss function jointly optimizes bounding box regression (localization) and classification, enabling balanced improvement in both detection accuracy and localization precision.

YOLO Applied to Segmentation Tasks

However, the standard YOLO model does not natively support segmentation. In many computer vision applications, image segmentation is indispensable—its goal being to partition an image into distinct regions, each corresponding to a specific object or background class. Consequently, adapting YOLO for segmentation usually requires architectural extensions or modifications.

After finishing “YOLO Segmentation Network: An Introduction to Object Detection and Image Segmentation”, take one minute to reflect:

• Are key concepts clearly distinguished?
• Are practical steps reproducible?
• Can you restate the conclusions in your own words?

YOLOv5 and Segmentation Networks

YOLOv5 is a pivotal version in the YOLO series. Building upon its strong object detection foundation, it introduces optional segmentation capabilities. Below are the fundamental steps to implement image segmentation using YOLOv5:

1. Dataset Preparation: Prepare a dataset containing segmentation annotations (e.g., the COCO dataset).
2. Model Selection: Choose YOLOv5 and configure it for segmentation mode.
3. Model Training: Train the model using the prepared dataset.
4. Inference Execution: Apply the trained model to perform segmentation on new images.

Example Code

Below is a basic code example demonstrating how to perform image segmentation using YOLOv5:

# Install YOLOv5
!git clone https://github.com/ultralytics/yolov5  # Clone the YOLOv5 repository
%cd yolov5
!pip install -r requirements.txt  # Install dependencies

import torch

# Load a pre-trained YOLOv5 segmentation model
segmentation_model = torch.hub.load('ultralytics/yolov5', 'yolov5s', pretrained=True)

# Load test image
img = 'test.jpg'  # Input image to be segmented

# Run segmentation inference
results = segmentation_model(img)

# Display results
results.show()
# Save results
results.save()

Interpreting Results

• The results.show() method displays the input image alongside detected objects and their corresponding segmentation masks.
• The results.save() method saves the annotated image—including bounding boxes and segmentation masks—to disk.

This example highlights two essential steps:

1. Loading a pre-trained YOLOv5 segmentation model via torch.hub.
2. Performing inference on an input image to obtain segmentation outputs.

When reviewing “YOLO Segmentation Network: An Introduction to Object Detection and Image Segmentation”, place key concepts, procedural steps, and observable outcomes side-by-side on a single page for effective consolidation.

When practicing “YOLO Segmentation Network: An Introduction to Object Detection and Image Segmentation”, explicitly document the input conditions, processing actions, and visible outputs together—facilitating efficient review and replication later.

Conclusion

In this article, we examined how the YOLO model can be applied to image segmentation—particularly through YOLOv5’s segmentation extension—to unify object detection and pixel-level segmentation in a single, efficient pipeline. This approach demonstrates how state-of-the-art detection techniques can be adapted to segmentation tasks, laying solid groundwork for further deep learning research.

In the next article, we will dive into YOLO’s source code to uncover its internal implementation details and optimization strategies—stay tuned!