0

Without GPU, using YOLO (darkflow), how feasible is it to detect objects using camera

Aasem
  • 11
  • 1
  • 1
    how do you define "realtime"? On a modern cpu, yolov3 needs about 300ms. Tiny yolo might be possible with much more fps. – Micka Apr 24 '19 at 11:45
  • Without any kind of neural network accelerators like [Google Coral](https://coral.withgoogle.com/) or [Intel Neural Compute Stick](https://software.intel.com/en-us/movidius-ncs)? If so, you should look at lightweight networks (Tiny YOLO, SqueezeNet, ...) and quantized networks (`int8`, [XNOR-Net:](https://pjreddie.com/media/files/papers/xnor.pdf), ...). – Catree Apr 24 '19 at 16:13

0 Answers0