search

cwq159 / PyTorch-Spiking-YOLOv3

A PyTorch implementation of Spiking-YOLOv3. Two branches are provided, based on two common PyTorch implementation of YOLOv3(ultralytics/yolov3 & eriklindernoren/PyTorch-YOLOv3), with support for Spiking-YOLOv3-Tiny at present.
GNU General Public License v3.0
207 stars 58 forks source link

snn output doesn't match the detection head #55

Open ThunderDavid opened 1 year ago

ThunderDavid commented 1 year ago

I use 'yolov3-tiny-mp2conv-mp1none-lk2relu-up2tconv' configuration to train a normal ANN model and reached a roughly 39% mAP in coco dataset, then I run 'ann_to_snn.py' script and have a transferred SNN and I got a huge .pth snn model file. Everything seems fine till now. But when I run the 'snn_evaluate' function in 'ann_to_snn.py' script, in output_snn1, output_snn2 = snn(data) # two branches, it goes wrong. I found that the snn(data) is a 19 elements tuble and I don't know how to let these 19 output match the output_snn1 and outpu_snn2. Can you help me? Yours faithfully

123456ZF commented 1 year ago

@DuanDexin It may be that the 30 series graphics card does not support ann_to_snn.py, so there will be an error, and I am experiencing the same problem.

123456ZF commented 1 year ago

@DuanDexin My friend ran successfully with 1650, 2060

Bender-L commented 1 year ago

You can try pytorch1.4 with python 3.7

123456ZF commented 1 year ago

@Bender-L Hello, how can I use the converted weights file of ann-snn for object detection, can it be used for detection?