search

qfgaohao / pytorch-ssd

MobileNetV1, MobileNetV2, VGG based SSD/SSD-lite implementation in Pytorch 1.0 / Pytorch 0.4. Out-of-box support for retraining on Open Images dataset. ONNX and Caffe2 support. Experiment Ideas like CoordConv.
https://medium.com/@smallfishbigsea/understand-ssd-and-implement-your-own-caa3232cd6ad
MIT License
1.4k stars 531 forks source link

How do we train on custom datasets #118

Open tiru1930 opened 4 years ago

tiru1930 commented 4 years ago

Hi

I have facemask dataset, i want to train ssd-mobilnet v1, I was able to train but my prediction are very bad.

for instance below is MAp

Average Precision Per-class: face: 0.0 face_mask: 1.937778212873579e-08

Average Precision Across All Classes:9.688891064367895e-09

So What steps i need to take care when i want to train on custom datasets.

  1. Do I need to change the priors ( anchor boxes)
  2. Can I use pre-trained SSD-mobilenetv1 or shall I start to using basenet alone.
  3. What kind of fine-tuning I can do to ?
thimabru1010 commented 4 years ago

I want to know too.

bijjuair commented 4 years ago

Did you get a solution? I am also facing the same problem.