hova88 / PointPillars_MultiHead_40FPS

A REAL-TIME 3D detection network [Pointpillars] compiled by CUDA/TensorRT/C++.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

how to train custom dataset in MultiHead

qiuhui1991 opened this issue · comments

No description provided.

the MULTIHEAD right now, Specifically, the postprocess.cu is written pretty ugly. There still has so many details (or paraments) that need to be adjusted manually. so, if you want to custom your own dataset or model framework, you have to read my ugly code. or just waiting for me to make it beautiful.

thank you very much for your reply

No description provided.

the MULTIHEAD right now, Specifically, the postprocess.cu is written pretty ugly. There still has so many details (or paraments) that need to be adjusted manually. so, if you want to custom your own dataset or model framework, you have to read my ugly code. or just waiting for me to make it beautiful.

i have complete custom dataset in single head, but i want to know how to transfor single head model to onnx, we need to change something in trans_backbone_multihead.py ?

No description provided.

the MULTIHEAD right now, Specifically, the postprocess.cu is written pretty ugly. There still has so many details (or paraments) that need to be adjusted manually. so, if you want to custom your own dataset or model framework, you have to read my ugly code. or just waiting for me to make it beautiful.

i have complete custom dataset in single head, but i want to know how to transfor single head model to onnx, we need to change something in trans_backbone_multihead.py ?

right. I think you need to modify trans_backbone_multihead.py and onnx_dense_head.py.

  1. try to define your singlehead framework in 'onnx_dense_head.py',
  2. use ‘trans_backbone_*.py` to do import/export stuff.

No description provided.

the MULTIHEAD right now, Specifically, the postprocess.cu is written pretty ugly. There still has so many details (or paraments) that need to be adjusted manually. so, if you want to custom your own dataset or model framework, you have to read my ugly code. or just waiting for me to make it beautiful.

i have complete custom dataset in single head, but i want to know how to transfor single head model to onnx, we need to change something in trans_backbone_multihead.py ?

right. I think you need to modify trans_backbone_multihead.py and onnx_dense_head.py.

1. try to define your singlehead framework in 'onnx_dense_head.py',

2. use ‘trans_backbone_*.py` to do import/export stuff.

i confused to define singlehead in onnx_dense_head.py, can you give me example for singlehead framework?

No description provided.

the MULTIHEAD right now, Specifically, the postprocess.cu is written pretty ugly. There still has so many details (or paraments) that need to be adjusted manually. so, if you want to custom your own dataset or model framework, you have to read my ugly code. or just waiting for me to make it beautiful.

i have complete custom dataset in single head, but i want to know how to transfor single head model to onnx, we need to change something in trans_backbone_multihead.py ?

right. I think you need to modify trans_backbone_multihead.py and onnx_dense_head.py.

1. try to define your singlehead framework in 'onnx_dense_head.py',

2. use ‘trans_backbone_*.py` to do import/export stuff.

i confused to define singlehead in onnx_dense_head.py, can you give me example for singlehead framework?

thank you for your project, i haved convert backbone_multihead.onnx and multihead_pfe.onnx succesfully in my custom dataset.

No description provided.

the MULTIHEAD right now, Specifically, the postprocess.cu is written pretty ugly. There still has so many details (or paraments) that need to be adjusted manually. so, if you want to custom your own dataset or model framework, you have to read my ugly code. or just waiting for me to make it beautiful.

i have complete custom dataset in single head, but i want to know how to transfor single head model to onnx, we need to change something in trans_backbone_multihead.py ?

right. I think you need to modify trans_backbone_multihead.py and onnx_dense_head.py.

1. try to define your singlehead framework in 'onnx_dense_head.py',

2. use ‘trans_backbone_*.py` to do import/export stuff.

i confused to define singlehead in onnx_dense_head.py, can you give me example for singlehead framework?

thank you for your project, i haved convert backbone_multihead.onnx and multihead_pfe.onnx succesfully in my custom dataset.

great! glad it has been useful to you. hope you enjoy your code.

Hi,

Can you please provide changes needed to be done to convert the multihead to singlehead?

No description provided.

the MULTIHEAD right now, Specifically, the postprocess.cu is written pretty ugly. There still has so many details (or paraments) that need to be adjusted manually. so, if you want to custom your own dataset or model framework, you have to read my ugly code. or just waiting for me to make it beautiful.

i have complete custom dataset in single head, but i want to know how to transfor single head model to onnx, we need to change something in trans_backbone_multihead.py ?

right. I think you need to modify trans_backbone_multihead.py and onnx_dense_head.py.

1. try to define your singlehead framework in 'onnx_dense_head.py',

2. use ‘trans_backbone_*.py` to do import/export stuff.

i confused to define singlehead in onnx_dense_head.py, can you give me example for singlehead framework?

thank you for your project, i haved convert backbone_multihead.onnx and multihead_pfe.onnx succesfully in my custom dataset.

Hi @qiuhui1991,
would you mind to share your onnx_dense_head.py?

commented

@hova88 @qiuhui1991 @Thomas-Kb @chasingw

Has your single header version been successfully implemented?

Would you mind to share your onnx_dense_head.py and trans_backbone_singlehead.py ?