Real Time Object Detection using YOLO V5

Real Time object detection is a technique that is used to recognize objects from video or photos. Over Years, there are many papers, libraries and models published for real time object detection. Among them You Only Look Once(YOLO) has dominated most of them. YOLO is one of the most accurate, realtime and versatile  object detection model. YOLO has gained a lot of popularity since its release.
Beginners Guide to Object Detection using YOLO V5
Object detection is one of the prime domain in computer vision. It plays an important role in various practical scenarios. Due to limitation of hardware, it is often necessary to sacrifice accuracy to ensure the infer speed of the detector in practice.  YOLO aims to implement an object detector that can balance effectiveness and efficiency so that it can be directly applied in actual application scenarios, rather than just propose a novel detection model.
From YOLO family YOLO V5 is ta latest release. It is proposed on 27 May 2020 by Glenn Jocher (founder of utralytics). YOLO V5 is written in pytorch and it is available to everyone through github. YOLO V5 is the fifth member of YOLO Family. Its predecessors are YOLO V1, YOLO V2, YOLO V3, and,  YOLO V4. YOLO V5 is very close to YOLO V4 however it collects most of its performance from pytorch training phenomena. The YOLO algorithms at first divides the given image into n*n grid. Then each grid is given the responsibility of object detection. Each box is given five parameters; the x & y coordinates, width and height of box and the confidence level of the probability of the object detection.
YOLO V5 has raised the standards for real time object detection. It already beats its competitors line EfficientDet and its previous versions.
Now let's Implement YOLO V5 in google collab:
Now, we are going to implement YOLO V5 in a very basic way. At first open google collab or jupyter notebook. The follow the following steps:
  1. At first, clone the github repository of YOLO V5 to your google colab environment. If you don't know how to do that use the command below:
!git clone # clone repo
    2. Then you need the install the required dependencies. The required dependencies are mentioned in the requirements.txt file inside the repository. You can install all the dependencies at once by using following command:
%cd yolov5 %pip install -qr requirements.txt # install dependencies
    3. Now lets import some modules to present our output in organized way. You can do this by using following command:
import torch from IPython.display import Image, clear_output
4. Now, there are some images inside runs/detect/exp/ this folder of the cloned repository. You can also use your custom image if you want. Now use to run inference on variety of sources and use it to detect objects in the given image. Use the following command:
!python --weights --img 640 --conf 0.25 --source data/images/ Image(filename='runs/detect/exp/bus.jpg', width=600)
    5. Our output looks like below: 

You can find the link to the colab file here.


We have now implemented the YOLO V5 object detection models and also seen it's popularity. It is the state of the art model for real time object detection. There are many other detailed tutorials over internet. If you want to go over them here are some tutorial and articles that I found really good:

Post a Comment