search

AlexeyAB / yolo2_light

Light version of convolutional neural network Yolo v3 & v2 for objects detection with a minimum of dependencies (INT8-inference, BIT1-XNOR-inference)
MIT License
302 stars 116 forks source link

Saving INT8 weights #44

Open willbattel opened 5 years ago

willbattel commented 5 years ago

I'd like to save the INT8 weights, to get the benefit of the smaller file size. I'm deploying YOLO to devices that are limited in disk space where going from 240MB to 60MB would be extremely helpful. Tiny YOLOv3 is not quite good enough for small objects, as much as we'd like to use it. The smaller file size is even more important to us than the 30% gain in speed, and we're okay with losing a little bit of accuracy.

I know several months ago you said you don't have plans to add the ability to save the INT8 weights. Is this still the case?

AlexeyAB commented 5 years ago

May be I will add it later.

abhigoku10 commented 5 years ago

@AlexeyAB is this functionality added ??

sicarioakki commented 5 years ago

@AlexeyAB Has this repo been updated to save the int8 weights functionality?

ambr89 commented 4 years ago

Hi @willbattel , did you be able to obtain INT8 weight? Anyone can do this?

ilpapds commented 1 year ago

Hi, I am sorry to ask. Is this functionality added. i am interested in saving yolov3 with int8 weights