SSGRL icon indicating copy to clipboard operation
SSGRL copied to clipboard

Published 20 hours ago • verified publisher icon HCPLab-SYSU

coco map is 0.037 in epoch15

Open valencebond opened this issue 5 years ago • 8 comments

i just run the code according to readme file and adjust it to pytorch1.1, python3.6. Is there something should be changed?

valencebond avatar Sep 20 '19 06:09 valencebond

I run it in the environment of PyTorch 0.4.1, Python 2.7. I updated one dataset interface in utils/transform.py, but I already set it in the script of each dataset. You can just update it and try again. In MS-COCO, I used cocoapi to load the data, have you used this? You can read the datasets/cocodataset.py to learn how I load those data, just a few sentences.

Mu-xsan avatar Sep 25 '19 03:09 Mu-xsan

i just run the code according to readme file and adjust it to pytorch1.1, python3.6. Is there something should be changed?

I adjust model to ResNet50/ResNet101. CoCo mAP is still 0.037 :(

KiAkize avatar Nov 25 '19 10:11 KiAkize

i just run the code according to readme file and adjust it to pytorch1.1, python3.6. Is there something should be changed?

I adjust model to ResNet50/ResNet101. CoCo mAP is still 0.037 :(

I got mAP 0.037 too.

phython96 avatar Nov 27 '19 15:11 phython96

i just run the code according to readme file and adjust it to pytorch1.1, python3.6. Is there something should be changed?

I adjust model to ResNet50/ResNet101. CoCo mAP is still 0.037 :(

I got mAP 0.037 too.

I got mAP 0.037 too.

zjhdxh avatar Dec 03 '19 08:12 zjhdxh

i just run the code according to readme file and adjust it to pytorch1.1, python3.6. Is there something should be changed?

I adjust model to ResNet50/ResNet101. CoCo mAP is still 0.037 :(

I got mAP 0.037 too.

I got mAP 0.037 too.

I solve the problem.You have to run with python 2.7 and pytorch 0.4.Then the mAP flies.

phython96 avatar Dec 03 '19 08:12 phython96

I got mAP 0.037 too.

i just run the code according to readme file and adjust it to pytorch1.1, python3.6. Is there something should be changed?

I adjust model to ResNet50/ResNet101. CoCo mAP is still 0.037 :(

I got mAP 0.037 too.

I got mAP 0.037 too.

I got mAP 0.037 too.

jasonseu avatar Dec 06 '19 06:12 jasonseu

I got mAP 0.037 too.

Yeah it's weird but true

student-petercai avatar Mar 31 '21 06:03 student-petercai

i just run the code according to readme file and adjust it to pytorch1.1, python3.6. Is there something should be changed?

How did you solve it in the end

NC122 avatar Sep 13 '21 11:09 NC122