Overview:
A global client in the transportation sector faced a significant challenge: processing 170,000 license plates accurately. The dataset presented multiple issues:
- Blurry Images: Poor image quality made it difficult to extract key details.
- Inconsistent Formats: License plates varied greatly in design, layout, and numbering conventions across regions.
- Varied Country Symbols: Flags or symbols indicating different countries were often obscured, missing, or highly inconsistent.
These challenges led to frequent errors in identifying license plate numbers, codes, and country symbols, ultimately impacting the client’s operational efficiency. To overcome these obstacles, the client partnered with Tasq.ai to implement a hybrid workflow combining AI, crowd validation, and manual BPO corrections.

The Problem:
- Blurry Images:
- Plates captured under poor lighting conditions or in motion led to misreading or incomplete data extraction.
- Important details, such as plate numbers, codes, and symbols, were often unclear or missing.
- Inconsistent Formats:
- Different regions employed unique license plate layouts, resulting in a lack of standardization.
- Variations in font, size, and alignment confused the recognition model.
- Country Symbol Identification:
- Country symbols, such as flags or emblems, were inconsistent in design and quality.
- Many images had symbols that were obscured, partially visible, or missing entirely.

The Solution:
To resolve these challenges, Tasq.ai deployed a three-layer hybrid workflow:
- AI Model Pre-Screening:
- The AI model scanned each plate for numbers, codes, and country symbols.
- Unclear or incomplete results were flagged and escalated for further validation.
- Crowd Validation:
- A diverse pool of trained contributors reviewed the AI’s results to ensure accuracy.
- This step provided a critical layer of quality assurance and minimized errors.
- BPO Fixes:
- For cases where both AI and the crowd could not provide a definitive answer, a specialized BPO team manually corrected the data.
- These corrections were fed back into the AI system to enhance future performance.
Workflow in Action:
- Step 1:License Plate Number Recognition
- AI extracted the plate number. Any uncertainties were reviewed by the crowd, and unresolved cases were escalated to the BPO team.
- Step 2:License Plate Code Validation
- AI identified the plate code. Format inconsistencies or errors were resolved through crowd validation and BPO corrections.
- Step 3:Country Symbol Identification
- AI detected the country symbol (e.g., flag or emblem). Any unclear or missing symbols followed the same process for validation and correction.
Results:
- Accurate Processing of 170,000 Plates: The hybrid workflow improved recognition accuracy, overcoming challenges like poor image quality and diverse formats.
- Error Reduction: Combining AI, crowd validation, and BPO improves the model to almost 100%.
- Model Enhancement: Corrected data was used to refine the AI model, enabling it to handle similar datasets more effectively in the future.
- Scalability: Tasq.ai’s platform processed the large dataset efficiently without compromising quality.
Why Tasq.ai?
This use case highlights Tasq.ai’s ability to deliver high-quality results for complex, large-scale challenges. By addressing issues like image clarity, format inconsistencies, and country symbol detection, Tasq.ai provided a seamless, scalable solution.
If you’re looking for a reliable, flexible approach to improve your AI workflows, Tasq.ai has the tools and expertise to make it happen. Contact us today for a demo!