cornell-zhang / bnn-fpga

Binarized Convolutional Neural Networks on Software-Programmable FPGAs
BSD 3-Clause "New" or "Revised" License
301 stars 112 forks source link

Binarized yolo #24

Closed Thilanka97 closed 4 years ago

Thilanka97 commented 5 years ago

@rzhao01 I am trying to run yolo on fpgas in a efficient way. I was thinking of maybe I can run binarized yolo on fpga using your work. will I be able to do this ? What changes do I need to make ? If I find a binarized weightsof yolo, then would it be possible ? please guide me.

Thanks in advanced !

rzhao01 commented 5 years ago

Running YOLO with this design would be challenging. My design is targeted for CIFAR-10 and uses a specialized first layer with 32x32 inputs. YOLO uses much larger inputs (448x448). YOLO also uses a combination of 7x7, 3x3, and 1x1 filters while the BNN in my design only uses 3x3 filters.

I would say you'd be better off looking for a different BNN accelerator which targets ImageNet and modify that.

Thilanka97 commented 5 years ago

@rzhao01 thanks alot for your reply. helps alot!