computerVision | common algorithms in computer vision | Computer Vision library

There are several key differences between the two. Form Recognizer's primary goal is to structure data from forms and other digitized documents for further processing. The key here is that Form Recognizer provides features that can help better contextualize the information that is read from said documents than just stand-alone optical character recognition. From the Form Recognizer documentation (emphasis mine):

Azure Form Recognizer is a cloud-based Azure Applied AI Service that uses machine-learning models to extract and analyze form fields, text, and tables from your documents. Form Recognizer analyzes your forms and documents, extracts text and data, maps field relationships as key-value pairs, and returns a structured JSON output. You quickly get accurate results that are tailored to your specific content without excessive manual intervention or extensive data science expertise. Use Form Recognizer to automate your data processing in applications and workflows, enhance data-driven strategies, and enrich document search capabilities.

On the other hand, Azure Computer Vision provides three distinct features. While the OCR tenet below describes something similar to Form Recognizer, it's more general-purpose in use in that it does not provide as robust contextualization of key/value pairs that Form Recognizer does. The service also provides higher-level AI functionality for processing images and video to identify people/celebrities, landmarks, and common objects in them (among others). From the Computer Vision documentation:

Service Description Optical Character Recognition (OCR) The Optical Character Recognition (OCR) service extracts text from images. You can use the new Read API to extract printed and handwritten text from photos and documents. It uses deep-learning-based models and works with text on a variety of surfaces and backgrounds. These include business documents, invoices, receipts, posters, business cards, letters, and whiteboards. The OCR APIs support extracting printed text in several languages... Image Analysis The Image Analysis service extracts many visual features from images, such as objects, faces, adult content, and auto-generated text descriptions. Follow the Image Analysis quickstart to get started. Spatial Analysis The Spatial Analysis service analyzes the presence and movement of people on a video feed and produces events that other systems can respond to. Install the Spatial Analysis container to get started.

At first glance, there is some overlap between the two, but upon further inspection there are clear delineations for the primary use cases for the two.

To help you understand, you need a bit of history of those services. Computer Vision API (and all "calling" SDKs, whether C#/.Net, Java, Python etc using these APIs) have moved frequently and it is sometimes hard to understand which SDK calls which version of the APIs.

Regarding optical character reading operations, there have been several operations:

See definition here was containing:

• OCR operation, a synchronous operation to recognize printed text
• Recognize Handwritten Text operation, an asynchronous operation for handwritten text (with "Get Handwritten Text Operation Result" operation to collect the result once completed)

See definition here. OCR was still there, but "Recognize Handwritten Text" was changed. So there were:

• OCR operation, a synchronous operation to recognize printed text
• Recognize Text operation, asynchronous (+ Get Recognize Text Operation Result to collect the result), accepting both printed or handwritten text (see mode input parameter)
• Batch Read File operation, asynchronous (+ "Get Read Operation Result" to collect the result), which was also processing PDF files whereas the other one were only accepting images. It was intended "for text-heavy documents"

Computer Vision 2.1 was similar in terms of operations.

See definition here. Main changes: Recognize Text and Batch Read File were "unified" into a Read operation, with models improvements. No more need to specify handwritten / printed for example (see link).

You can't replace resnet50's fc with a convolutional network. The output of resnet's feature extractor is a CNN which outputs a flat 2048-long tensor, as such the layers following it should be fully connected layers.

Please use the analyze image API, we are able to see response as shown below.

Here is quick-start sample using SDK and REST API.

The evaluate() function here doesn't calculate any loss. And look at how the loss is calculate in train_one_epoch() here, you actually need model to be in train mode. And make it like the train_one_epoch() except without updating the weight, like

Following a suggestion here, I discovered some of my images did not contain any bounding boxes and was causing this error:

I found you can use .get_layer() twice to acces layers inside functional densenet model embebeed in the "main" model.

In this case I can use model.get_layer('densenet121').summary() to check all thje layer inside the embebeed model, and then use them with this code: model.get_layer('densenet121').get_layer('xxxxx')

Pls make sure that you have created a current type of cognitive service, I recommend you to create All Cognitive Services just as below:

You can follow this doc to create it(Multi-service resource).

I did some test on my side by this service and everything works for me as expected :

My local test image:

Result:

The Computer Vision Read API is Azure's latest OCR technology that handles large images and multi-page documents as inputs and extracts printed text in Dutch, English, French, German, Italian, Portuguese, and Spanish. It also includes support for handwritten OCR in English, digits, and currency symbols from images and multi-page PDF documents. It's optimized to extract text from text-heavy images and multi-page PDF documents with mixed languages. It supports detecting both printed and handwritten text in the same image or document (for English only).

Here is the doc for release update in computer vision.