midasklr
midasklr
add write permission to '/usr/local/lib/python3.6/dist-packages/': sudo chmod a+x /usr/local/lib/python3.6/dist-packages/
> 我用的icdar2015数据集，loss不为0，但是精度和召回率一直都是0，有解决方法吗 怎么样？能正常训练识别了不？
> yeah, there are some problems, run demo.py, still happen no boxes to NMS, but can got mAP 0.5439399966. and training, the CPU cost is very hight, even over 85%...
I train efficientdet d0 on voc0712, i didnt get attention with cpu cost, have u test on voc2007? whats the mAP u got? mine is as: AP for aeroplane =...
> As I read paper, efficient d0 size should be 3.8MB. > Why d0 checkpoint weight included on git has 40MB size ?? I guess 3.8M, which is 3.9M in...
> > > As I read paper, efficient d0 size should be 3.8MB. > > > Why d0 checkpoint weight included on git has 40MB size ?? > > >...
> I have a question which might need your help or guidance on this. I am new to this. I followed your object detection with yolov5 to implement in android....
`new_state_dict = OrderedDict() for k, v in checkpoint['state_dict'].items(): name = k[7:] # remove module. new_state_dict[name] = v model.load_state_dict(new_state_dict)`
EAST TRAIN Epoch: [555][40/125] LR 0.0007339040223999998 Loss 0.0028 Avg Loss 0.0040) EAST TRAIN Epoch: [555][50/125] LR 0.0007339040223999998 Loss 0.0021 Avg Loss 0.0041) EAST TRAIN Epoch: [555][60/125] LR 0.0007339040223999998 Loss 0.0019...
> 标注之后需要用里面的脚本split，预训练模型可以从mxnet官方下载。我的模型是我自己的训练集训练的。 大佬能否更新下数据集链接，链接已经失效啦，感谢