Computer vision and document recognition
Recognising objects in photos and video, segmenting images, extracting data from scans, forms and source documents. We build the models into the process so the results go straight into your systems.
Who it is for
Companies where people copy data out of documents by hand, check photos and video, or count objects — in the warehouse, in production, in logistics or in document workflows.
What the client gets
A recognition model built into your process, with quality checked on your own images and documents. We hand over the training and deployment code and a description of the data and the limits.
Object recognition
Detecting and classifying objects in photos and video: goods, defects, vehicles and other objects in frame.
Image segmentation
Precise outlining of image regions where the boundaries and area of an object matter.
Document recognition
OCR and field extraction from scans, forms, invoices and source documents.
Training on your data
Labelling examples and fine-tuning models for your shooting conditions and document types.
Integration and review
Recognition results go into accounting, the CRM or document workflow, and doubtful cases go to a person for review.
Questions
Will off-the-shelf models do?
Sometimes — we check that first. If the shooting conditions or the documents are specific, we fine-tune the model on labelled examples from your process.
How accurate will recognition be?
Accuracy depends on the quality of the images and the data, so we do not guarantee metrics. We measure quality on your examples and plan human review for doubtful cases.
Can recognition run without the cloud?
Yes. The model can be deployed on your servers if images and documents must not leave your infrastructure.