ID scanning solutions developer OCR Studio has announced the launch of IIRDoc-Net, a neural network architecture that can recognise IDs with 32% less compute.
Available on mobile phones, tablets, and desktops, the newly rolled-out AI technology is currently leveraged by OCR Studio’s clients across several sectors, including banking, retail, and healthcare.
Using AI for ID recognition
Automated identity document recognition is widely utilised in industries such as banking, logistics, and border control, and the current shift toward mobile platforms requires these technologies to be performed even on low-budget devices. Additionally, recognition models need to have an extended receptive field to maintain calculation accuracy and to handle document images’ projective distortions, noise, motion blur, and poor lighting. According to OCR Studio, most of the models with wide receptive fields are too heavy to provide a real-time document analysis on a mobile device. To mitigate these issues, the company launched the IIRDoc-Net solution, which integrates a block of learnable Infinite Impulse Response (IIR) filters, enabling the receptive field to be adjusted by integrating both current inputs and previous outputs.
Following evaluations on the MIDV-500 and MIDV-2019 datasets, OCR Studio’s solution offered a 32% decrease in computation required for ID recognition when compared to other approaches. The technology has been implemented in the latest version of OCR ID-scan, used by government, fintech, HR and other services for document-processing automation.
Furthermore, OCR Studio intends to release a paper in the electronic proceedings of the International Society of Optics and Photonics (SPIE) about the IIRDoc-Net technology, which comes as the company’s debut at an international scientific conference of this kind.
Latest news from OCR Studio
In addition to launching the IIRDoc-Net solution, OCR Studio recently introduced its Arabic ID document scanner, applicable on the web page, a capability developed into its browser-based ID scanner. The AI system was set to read Arabic passports and ID cards in Chrome, Safari, Firefox, and other browsers without transmitting the data from the end-user’s device. Developed specifically for corporate clients across the MENA region, the ID scanner was set to deliver a compliant and reliable tool for finance, government, healthcare, and retail companies.
