{"id":8516,"date":"2024-01-03T06:00:27","date_gmt":"2024-01-03T14:00:27","guid":{"rendered":"https:\/\/live-cometml.pantheonsite.io\/?p=8516"},"modified":"2025-04-24T17:03:38","modified_gmt":"2025-04-24T17:03:38","slug":"build-and-monitor-an-object-detection-model-in-5-steps-using-comet","status":"publish","type":"post","link":"https:\/\/www.comet.com\/site\/blog\/build-and-monitor-an-object-detection-model-in-5-steps-using-comet\/","title":{"rendered":"Build and Monitor an Object Detection Model in 5 Steps Using Comet"},"content":{"rendered":"\n

In this project guide, we will dive into object detection by creating a custom object detection model and monitoring its accuracy metrics and hyperparameters in five simple steps!<\/p>\n\n\n\n

It doesn’t even matter if your machine sucks, let’s go!<\/p>\n\n\n\n

Dataset<\/h2>\n\n\n\n

We will use this dataset<\/a> from Google Drive to train the model in this project. The dataset contains two folders: train<\/strong> and test, which both<\/strong> contain subfolders, images,<\/strong> and labels. As<\/strong> the name implies here, the images folder includes the training and validation images, while the labels<\/strong> have theannotation for the respective images. <\/strong>We will be uploading this file to our Google Colab soon. Get ready!<\/p>\n\n\n\n

Coding Environment<\/h2>\n\n\n\n

We will be using Google Colab as our coding environment for this project.<\/p>\n\n\n\n

\n

Ensure to turn your Runtime to GPU when using Google Colab.<\/p>\n<\/blockquote>\n\n\n\n

Step 1: Clone the YoloV7 repository from GitHub<\/h2>\n\n\n\n

We need to clone the YoloV7 folder from the official GitHub repository:<\/p>\n\n\n\n

! git clone https:\/\/github.com\/WongKinYiu\/yolov7.git<\/span><\/pre>\n\n\n\n
Code Output<\/h2>\n\n\n\n
\"Cloning
Cloning the official Yolov7 repository<\/figcaption><\/figure>\n\n\n\n
<\/p>\n\n\n\n
From the code output above, it can be seen that we can create a folder “yolov7,<\/strong>” which contains all the subfolders, scripts, configuration files, and many more required to run object detection.<\/p>\n\n\n\n
Step 2: Install All The Dependencies Needed to Run Your Object Detection<\/h2>\n\n\n\n
We must install the dependencies provided in the requirement.txt file<\/strong> inside the yolov7<\/strong> folder. But first, you need to ensure that you are inside the yolov7 <\/strong>folder before running the requirement.txt <\/strong>file. To do this, run the command below:<\/p>\n\n\n\n
cd \/content\/yolov7<\/span><\/pre>\n\n\n\n
This will return the output below:<\/p>\n\n\n\n
\"code<\/figure>\n\n\n\n
Now run the requirements.txt<\/strong> files:<\/p>\n\n\n\n
pip install -r requirements.txt<\/span><\/pre>\n\n\n\n
Code Output<\/h2>\n\n\n\n
\"Installing
Installing requirements<\/figcaption><\/figure>\n\n\n\n
Step 3: Prepare and Upload Your Data<\/h2>\n\n\n\n
Now proceed inside the yolov7<\/strong> folder and locate the data<\/strong> folder. Open it and create two folders, test<\/strong> and train<\/strong>. Create subfolders, images, and labels inside each of the two folders.<\/p>\n\n\n\n
The figure below shows the newly created folders.<\/p>\n\n\n\n
\"Data
Data folder showing train and test datasets<\/figcaption><\/figure>\n\n\n\n
<\/p>\n\n\n\n
Bravo! Now, go and download the dataset<\/a> we will be using for this project into your local machine.<\/p>\n\n\n\n
The idea is that the folder structure and all its contents must be uploaded into its corresponding folder in Google Colab. This means that for every image in the test<\/strong> folder on your local machine, you select all and drag and drop them inside the images<\/strong> folder of the test<\/strong> folder in Google Colab. You also do this for the labels. When you are done with the test <\/strong>folder, repeat the same for the train<\/strong> folders; select all, drag, and drop all the images and labels in the train folder in your local machine into the images and labels folders in the train folder here in Google Colab.<\/p>\n\n\n\n
To confirm you uploaded the files correctly and successfully, run the code below:<\/p>\n\n\n\n
import<\/span> os\ntest_images = \"\/content\/yolov7\/data\/test\/images\"<\/span>\ntest_labels = \"\/content\/yolov7\/data\/test\/labels\"<\/span>\ntrain_images = \"\/content\/yolov7\/data\/train\/images\"<\/span>\ntrain_labels = \"\/content\/yolov7\/data\/train\/labels\"<\/span>\n\nprint<\/span>(\"Number of images and labels inside the test folder:\"<\/span>, len<\/span>(os.listdir(test_images)) + len<\/span>(os.listdir(test_labels)))\nprint<\/span>(\"Number of images and labels inside the train folder:\"<\/span>, len<\/span>(os.listdir(train_images)) + len<\/span>(os.listdir(train_labels)))\n<\/span><\/pre>\n\n\n\n
Code Output:<\/p>\n\n\n\n
\"code<\/figure>\n\n\n\n
From the code output above, the number of training images alongside their corresponding labels is 72, while that of the validation is 32.<\/p>\n\n\n\n
Step 4: Set Up Your Configuration Scripts<\/h2>\n\n\n\n
Here is where it might get tricky. But sit tight; it is as simple as ABC!<\/p>\n\n\n\n
Edit the coco.yaml file<\/h2>\n\n\n\n
Inside your yolov7 > data<\/strong> folder, there is a file named coco.yaml <\/strong>file. Open it and incorporate the following edits:<\/p>\n\n\n\n