daybreak-u / centernet_mobilenetv2_ncnn Goto Github PK

i want compile this code in arm，there is something wrong about this library

请问下作者是使用的哪个版本的centernet训练的呀？
https://github.com/xingyizhou/CenterNet 是这个吗？

大神，您好，非常感谢分享工程，这个工程对学习centernet部署非常有帮助。在pytorch 1.2.0环境霞用mobilenetv2_10骨干网训练模型运行成功后，尝试把模型转换成onnx格式，遇到错误。pytorch新手，期望得到您的指点！
`
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torchvision import datasets, transforms
from torch.autograd import Variable
import torch.onnx as torch_onnx

import _init_paths

from detectors.detector_factory import detector_factory
from models.model import create_model, load_model
from opts import opts

heads = {'hm': 2, 'wh': 2, 'hps': 8, 'hm_hp': 4}
model = create_model('mobilenetv2_10', heads, 64)
model = load_model(model, '../model_best.pth')
torch.save(model, 'mobilnetv2_10.pth')

print(model)

input_shape = (3, 512, 384)
dump_input = Variable(torch.randn(1, *input_shape))
output = torch_onnx.export(model, dump_input, 'mobilnetv2_10.onnx', verbose=False)
print('Export of torch model completed!')
`

错误信息

Traceback (most recent call last): File "/train-data/CenterNet/src/tools/export_onnx.py", line 24, in <module> output = torch_onnx.export(model, dump_input, 'model.onnx', verbose=False) File "/home/link/.conda/envs/CenterNet/lib/python3.6/site-packages/torch/onnx/__init__.py", line 132, in export strip_doc_string, dynamic_axes) File "/home/link/.conda/envs/CenterNet/lib/python3.6/site-packages/torch/onnx/utils.py", line 64, in export example_outputs=example_outputs, strip_doc_string=strip_doc_string, dynamic_axes=dynamic_axes) File "/home/link/.conda/envs/CenterNet/lib/python3.6/site-packages/torch/onnx/utils.py", line 329, in _export _retain_param_name, do_constant_folding) File "/home/link/.conda/envs/CenterNet/lib/python3.6/site-packages/torch/onnx/utils.py", line 213, in _model_to_graph graph, torch_out = _trace_and_get_graph_from_model(model, args, training) File "/home/link/.conda/envs/CenterNet/lib/python3.6/site-packages/torch/onnx/utils.py", line 171, in _trace_and_get_graph_from_model trace, torch_out = torch.jit.get_trace_graph(model, args, _force_outplace=True) File "/home/link/.conda/envs/CenterNet/lib/python3.6/site-packages/torch/jit/__init__.py", line 256, in get_trace_graph return LegacyTracedModule(f, _force_outplace, return_inputs)(*args, **kwargs) File "/home/link/.conda/envs/CenterNet/lib/python3.6/site-packages/torch/nn/modules/module.py", line 547, in __call__ result = self.forward(*input, **kwargs) File "/home/link/.conda/envs/CenterNet/lib/python3.6/site-packages/torch/jit/__init__.py", line 324, in forward out_vars, _ = _flatten(out) RuntimeError: Only tuples, lists and Variables supported as JIT inputs, but got dict

您好，感谢分享此工程，请问下您是如何将pytorch模型转换为ncnn模型的呢？采用的是onnx还是其他什么框架呢？谢谢！

大神，论文里用maxpool2求极值计算目标中心点的，本项目的genIds函数里判断超过阈值的score输出“候选”目标，然后再NMS，有研究过这两者之间的差异吗?

def _nms(heat, kernel=3):
pad = (kernel - 1) // 2
hmax = nn.functional.max_pool2d( heat, (kernel, kernel), stride=1, padding=pad)
keep = (hmax == heat).float()
return heat * keep

A very good job, thanks to open source.