Optical Character Recognition
Custom OCR
Our OCR model extracts English texts from scene images with more than 95% accuracy. This can be used to detect bib numbers from race images, detect labels from social media images, detect information from identity cards and more.
Benefits
- Developed to run on standard local hardware and on cloud instance as well.
- Can be trained on custom dataset for special application
- Inference time 0.5 to 5 seconds with 500 x 300 pixel images
- Integration through API is possible
Market
OCR is high in demand for all domains from photography, retail, e-commerce, social media and more.