StradVision achieved FOUR No.1 performance in world-class tech evaluations!
|Low-level Feature Learning||Deep Learning||StradVision's Agile Learning|
|Train||Fast, need reasonable amount of data||Very slow, Need huge dataset||Fast, need reasonable amount of data|
|Run||Fast (Can be run @ Wearable, IOT)||Slow (Need GPU)||Fast (Can be run @ Wearable, IOT)|
|BMT result||Defeated by deep learning in many BMT's||Very good as follows:
Object Recognition No.1
Face Recognition No.1
|Very good as follows:
Text Detection No.1 (EN & CN & Kannada)
Pedestrian Detection No.1
|Debugging||Doable||Very difficult||Offers visual debugging tool|
|Coding vs Data||Data is need.
Requires a lot of a prior knowledge, and a lot of trial and error.
|Data is everything, no prior knowledge||Data is everything. No prior knowledge is required.
And yet, prior knowledge, if available,
can be naturally incorporated into learning.
|Domain||Detection, Recognition, Classification, Anything||Detection (yet very slow), Recognition, Classification||Detection, Recognition, Classification|