We are use TDA4 TIDL to 4 cameras inference. How can I use batch inference to reduce inference time?
I search the forum and find that in SDK8.0, i can do it with numBatches para.
Can I achieve the same effect in SDK7.1 with other alternative methods? For example, can I concate the two images in channel dim and use the groupconv(Conv2d(groups=2 in pytorch)) to achieve similar result? Or is there other methods.
