AI-Boosted QR & Barcode Scanning

Barcode scanners are essential tools for logistics and cargo companies, enabling them to manage their operations more effectively, reduce errors, improve customer satisfaction, and enhance overall productivity. Barcode scanners are used to track the location and condition of these assets, making it easier to locate and maintain them.

While barcode scanners offer significant advantages to logistics and cargo companies, there are also challenges associated with their use in this sector:

• Barcode Quality: The quality of barcodes can vary significantly. Poorly printed or damaged barcodes can be difficult to scan accurately, leading to errors in data collection and inventory management.
• Environmental Factors: Harsh environmental conditions, such as extreme temperatures, humidity, dust, and exposure to the elements, can affect the performance and durability of barcode scanners.
• Scanning Range: Depending on the specific barcode technology used (e.g., 1D or 2D barcodes), scanners may have limited scanning ranges. Ensuring that scanners can read barcodes on items located in high shelves or distant areas of a warehouse can be challenging.
• Costs: While barcode scanners are essential tools, the initial cost of purchasing the equipment, along with ongoing costs for maintenance and upgrades, can be a significant investment for logistics companies.

Barcodes can be susceptible to damage during the movement of goods and cargo, especially if they are not properly protected or if they come into contact with rough handling or adverse environmental conditions. Here are some common types of damage that can occur to barcodes during transportation:

• Torn or Scratched Labels: Barcode labels on packages or products can get torn, scratched, or scuffed during handling or if they come into contact with other items or surfaces. This can make it difficult for barcode scanners to read the barcode accurately.
• Fading or Sun Exposure: Barcodes printed on labels can fade over time, especially if exposed to direct sunlight or UV radiation during transportation. Faded barcodes may become unreadable.
• Smudging or Stains: Spills or exposure to liquids during transport can cause ink-based barcodes to smudge or become stained. This can obscure the barcode and hinder scanning.
• Tape or Adhesive Residue: Sometimes, additional tape or labels are applied over the original barcode labels for various reasons. When these additional labels are removed, they can leave behind residue that may interfere with barcode scanning.
• Label Peeling: Barcode labels may start to peel off or detach from packages due to changes in temperature and humidity or poor adhesive quality. Partially detached labels can be challenging to scan.
• Crinkled Labels: If packages are subjected to compression or impact, barcode labels can become crinkled or wrinkled, making it difficult for scanners to read the barcode accurately.
• Barcodes on Irregular Surfaces: Barcodes applied to irregularly shaped or uneven surfaces may not lay flat, which can result in distortion or skewing of the barcode. This can affect scanning reliability.
• Damage to Barcode Printers: Barcode labels are typically printed using thermal printers. If the printer is damaged during transportation, it can lead to printing defects on barcode labels, making them unreadable.
• Exposure to Extreme Temperatures: Extreme temperatures, both hot and cold, can affect the adhesive properties of barcode labels and the readability of printed barcodes. Labels may not adhere well or may become brittle and crack.
• Abrasion: Barcodes can be abraded if they come into contact with abrasive surfaces or materials during handling or transportation, which can result in loss of barcode clarity.

Our project aims to leverage deep learning technology for the restoration of damaged barcodes. The core objectives of our project are as follows:

• Cost-Efficient Hardware: To ensure cost-effectiveness, we will employ affordable scanner hardware, specifically the Oak-1 camera.
• Compact and Efficient Model: We intend to develop a compact and efficient deep learning model that can efficiently run on the Oak camera's Vision Processing Unit (VPU).
• Dataset Creation: We will curate a comprehensive dataset, including images containing corrupted or damaged barcodes. This dataset will serve as the foundation for training our deep learning model.
• Generative Deep Learning: Our approach involves the use of generative deep learning techniques, particularly image-to-image translation, to restore the integrity of damaged barcodes. The model will learn to transform input images, featuring corrupted barcodes, into images displaying clean and restored barcodes.
• Input and Output: The input to our model will consist of images containing corrupted barcodes, while the model's output will be images showcasing the successfully restored barcodes.
• Implementation: We plan to deploy the final model on the AI accelerator integrated into the Oak-1 camera. This ensures real-time barcode restoration directly on the camera, enhancing the efficiency and applicability of our solution.

Presented below are preliminary results from our trained model executed on the Oak-1 camera, leveraging the Intel VPU accelerator.

In summary, our project seeks to employ deep learning, cost-effective hardware, and generative image restoration techniques to rectify damaged barcodes. The resulting model will run efficiently on the Oak-1 camera's hardware, providing a practical and real-time solution for barcode restoration.