Learn how to use VisionHowl to buy, sell, and create computer vision datasets
VisionHowl is a platform dedicated to buying, selling, and creating computer vision datasets. Whether you're a researcher looking for high-quality data, a company with proprietary datasets to monetize, or a developer building computer vision applications, VisionHowl provides the tools and marketplace to meet your needs.
Our platform supports various dataset formats, quality verification, and format conversion to ensure you can easily integrate the datasets with your existing workflows.
To use VisionHowl, you'll need to create an account. Registration is simple and only requires basic information.
Click on the "Register" button in the top-right corner of the navigation bar.
Enter your username, email address, and password. Make sure your password is secure.
After registration, you'll receive a verification email. Click the link in the email to verify your account.
Once verified, you can log in and complete your profile with additional information.
Once registered, you can log in to access your account.
Click on the "Login" button in the top-right corner of the navigation bar.
Enter your username and password.
After logging in, you'll be redirected to your dashboard where you can manage your datasets and account.
Your dashboard is the central hub for managing your datasets and account. Here's what you'll find:
At the top of your dashboard, you'll see key statistics:
Your datasets are organized into sections based on their status:
The sidebar provides quick access to common actions:
Additional tools to help you manage your datasets:
The dataset catalog is the main page for browsing available datasets. You can access it by clicking on "Datasets" in the navigation bar.
The catalog displays datasets in a grid layout with key information:
VisionHowl provides several ways to find the datasets you need:
Use the search bar at the top of the dataset catalog to search by keywords. This will match against dataset names, descriptions, and tags.
Use the filters in the sidebar to narrow down results by:
Sort the results by:
Clicking on a dataset in the catalog will take you to its detail page, where you can find comprehensive information:
Each dataset has a quality score based on:
To purchase a dataset, you first need to add it to your cart:
Find a dataset you're interested in and click on it to view its details.
On the dataset detail page, click the "Add to Cart" button. If the dataset is free, you'll see a "Download" button instead.
Click on the cart icon in the navigation bar to view your cart and proceed to checkout.
From the cart page, you can:
VisionHowl uses Stripe for secure payment processing. Here's how to complete your purchase:
Make sure all the datasets you want to purchase are in your cart.
This will take you to the Stripe checkout page.
Enter your credit card details or use another payment method supported by Stripe.
After your payment is processed, you'll be redirected back to VisionHowl.
You'll see a success message and receive a purchase receipt by email.
After purchasing a dataset, you can download it from your dashboard:
Click on "My Dashboard" in the navigation bar.
Scroll down to find the section showing your purchased datasets.
Click the "Download" button next to the dataset you want to download.
VisionHowl allows you to convert datasets between different formats:
Click the "Download" button for a dataset you've purchased.
You'll see a format selection page. Choose the format you want to convert to.
The system will convert the dataset to your selected format. This may take some time for large datasets.
Once the conversion is complete, the download will start automatically.
VisionHowl currently supports the following format conversions:
| Source Format | Target Formats |
|---|---|
| COCO | YOLO, Pascal VOC |
| YOLO | COCO, Pascal VOC |
| Pascal VOC | COCO, YOLO |
VisionHowl allows you to upload and sell your own datasets:
Click on "Upload Dataset" in your dashboard or navigation bar.
Enter the name, description, price, and other details about your dataset.
Upload your dataset files in a supported format (ZIP, TAR, etc.).
Upload sample images that showcase your dataset.
Submit your dataset for review by the VisionHowl team.
Metadata templates help you provide consistent and complete information about your datasets:
Click on "Metadata Templates" in your dashboard.
Click "Create New Template" and define the fields you want to include.
When uploading a dataset, select your template from the dropdown menu.
Complete all the fields defined in the template.
VisionHowl evaluates datasets based on several quality criteria:
How thoroughly you've documented your dataset:
The quality and representativeness of your preview images:
Whether your dataset passes format validation checks:
The quality of included documentation:
To improve your dataset's quality score:
All datasets go through a review process before being published:
After uploading your dataset, click "Submit for Review".
The VisionHowl team reviews your dataset for quality, content, and compliance.
You'll receive a notification when your dataset is approved or rejected.
Approved datasets are published in the marketplace.
VisionHowl allows you to maintain multiple versions of your datasets:
From your dataset detail page, click "Create New Version".
Upload the new version of your dataset files.
Provide a description of what changed in this version.
New versions also go through the review process.
VisionHowl provides detailed statistics for your datasets:
Track how many times your dataset detail page has been viewed.
See how many times your dataset has been downloaded.
Track how many times your dataset has been purchased.
See the percentage of viewers who purchase your dataset.
To view your dataset statistics:
Click on "My Dashboard" in the navigation bar.
Locate the dataset you want to view statistics for.
Click the "Stats" button next to the dataset.
You can update your profile information at any time:
Click on your username in the top-right corner and select "Profile".
Update your name, bio, profile picture, and other details.
Click "Save Changes" to update your profile.
Email verification is required to use all features of VisionHowl:
When you register, a verification email is sent to your email address.
Open the verification email and click the verification link.
You'll see a confirmation message that your email has been verified.
To change your password:
Click on your username in the top-right corner and select "Profile".
Click the "Change Password" button.
Enter your current password and your new password twice.
Click "Change Password" to update your password.
If you forget your password:
Click on "Login" in the navigation bar.
Click the "Forgot Password" link below the login form.
Enter the email address associated with your account.
Open the password reset email and click the reset link.
Enter your new password twice and click "Reset Password".
VisionHowl supports several common dataset formats:
A popular format for object detection, segmentation, and keypoint detection:
A simple format optimized for real-time object detection:
A format used in the PASCAL Visual Object Classes challenge:
Follow these best practices to create valuable datasets:
For a comprehensive guide on creating high-quality datasets, check out our detailed Dataset Guidelines.
Problem: Dataset upload fails or times out.
Solution: Try uploading smaller files (under 1GB), check your internet connection, or compress your dataset more efficiently.
Problem: Format conversion fails or produces incorrect results.
Solution: Ensure your source dataset is in a valid format, check for missing files or annotations, and try converting to a different format.
Problem: Payment fails or you're charged but don't receive access to the dataset.
Solution: Check your payment method, contact your bank if necessary, and reach out to VisionHowl support if you were charged but don't have access.
Problem: Can't log in or access your account.
Solution: Reset your password, check if your email is verified, and ensure you're using the correct username/email.
If you encounter issues not covered in this documentation: