The problem We want it that way... solves

Here's how our illegal hoardings detection project can benefit people and make existing tasks easier/safer:

Streamlining Urban Management

Our solution can significantly streamline the process of identifying and addressing illegal hoardings and billboards for urban management authorities. By automating the detection process, it eliminates the need for manual inspections and human error, making the task more efficient and cost-effective.

Improving Public Safety

Illegal hoardings can pose various safety risks, such as obstructing visibility on roads, promoting unsafe practices, or being structurally unsound. Our project helps identify these hazards promptly, allowing authorities to take prompt action and mitigate potential dangers, ultimately improving public safety.

Enhancing Cityscape Aesthetics

Unregulated hoardings can significantly detract from a city's visual appeal and character. Our solution aids in maintaining a clean and visually appealing urban landscape by quickly identifying and enabling the removal of unsightly and illegal billboards.

Enforcing Advertising Regulations

Many cities have regulations governing the content, size, and placement of outdoor advertising. Our project ensures these rules are upheld by detecting violations, such as inappropriate or offensive content, excessive sizes, or unauthorized locations.

Empowering Community Engagement

With our solution, citizens can actively participate in reporting illegal hoardings by submitting photographs through a dedicated app or platform. This citizen-driven approach fosters community engagement and enables a collaborative effort in maintaining a well-regulated urban environment.

Integrating with Existing Systems

Our project can seamlessly integrate with existing urban management systems and workflows. The detection results can be fed into existing databases, ticketing systems, or mobile apps, enabling a cohesive and efficient response to identified violations.

Challenges we ran into

During the development of our illegal hoardings detection project, we encountered a few significant hurdles that challenged our team's problem-solving skills. One particular issue that stands out is related to the language detection component.

The Kannada Language Hurdle

One of the key aspects of our project was to assess the language used on the hoardings, especially the presence and prominence of the local Kannada language. However, we quickly realized that finding reliable Optical Character Recognition (OCR) solutions for the Kannada script was a daunting task.

Initially, we attempted to leverage existing open-source OCR engines like Tesseract, but their performance on Kannada text was subpar. The engines struggled to accurately recognize and extract the Kannada characters from the images, leading to incomplete results.

The Solution: Google Gemini API and Prompt Engineering

After extensive research and experimentation, we stumbled upon Google's Gemini API, which offered language detection capabilities. While not designed specifically for OCR tasks, we realized that we could leverage the API in conjunction with prompt engineering techniques to estimate the language composition of the hoardings.

Our approach involved uploading the hoarding images to the Gemini API and carefully crafting prompts to extract language-related information. Although the API could not provide precise percentages of language usage, it allowed us to determine whether Kannada text was present on a given hoarding or not.