[echelon]

Are there any benchmarks (speed, accuracy, etc.) for non-OCR use cases? I want to label images and videos, but don't really care about text.