Key learnings when using Visual Perception skills

[AlexanderMarlow]

Visual Perception skills let your digital team operate with text the same way people do. Actions such as reading, interpreting human language and understanding screens, data and documents, bringing processes like real-time object and image identification to life.

We would like to know what special considerations or key learnings you have when using these skills. Please comment below if you have any fun or interesting stories as well!

------------------------------
Alexander Marlowe
Blue Prism
Blue Prism
------------------------------

[MattPalmer]

We've been lucky to have access to AWS Rekognition and MS Computer Vision in addition to the Tesseract engine included with BluePrism.  I've noticed in different scenarios one of the Vision tools might work better than the others and so its been good to have options to compare the image against all of the tools.  I've also built a custom action that allows us to change the size of the image that has helped increase the accuracy of what the Bot is attempting to see (usually increasing in size, but sometimes decreasing) .  I've not yet built any additional image actions, but have thought eventually we might need actions to change an image to grayscale or sharpen an image might also assist in increasing accuracy.

------------------------------
Matt Palmer
Business Process Specialist
Chesapeake Energy
America/Chicago
------------------------------

[AlexanderMarlow]

Thanks, Matt for sharing your insights. You mentioned that some tools work better than others when used in different scenarios, can you please detail the scenarios you have encountered? Also, why do you think these tools had different results? This is very interesting, we look forward to learning more from your experience!

Thanks.

------------------------------
Alexander Marlowe
Blue Prism
Blue Prism
------------------------------