| 字段名称 | 类型 | 描述 |
|---|---|---|
| filename | string | 图片名称 |
| X1 | int | 左上角X坐标 |
| Y1 | int | 左上角Y坐标 |
| X2 | int | 右上角X坐标 |
| Y2 | int | 右上角Y坐标 |
| X3 | int | 右下角X坐标 |
| Y3 | int | 右下角Y坐标 |
| X4 | int | 左下角X坐标 |
| Y4 | int | 左下角Y坐标 |
| type | int | 交通标示对应的编号(具体看下表) |
| 类型 | 对应的编号 |
|---|---|
| 停车场 | 1 |
| 停车让行 | 2 |
| 右侧行驶 | 3 |
| 向左和向右转弯 | 4 |
| 大客车通行 | 5 |
| 左侧行驶 | 6 |
| 慢行 | 7 |
| 机动车直行和右转弯 | 8 |
| 注意行人 | 9 |
| 环岛行驶 | 10 |
| 直行和右转弯 | 11 |
| 禁止大客车通行 | 12 |
| 禁止摩托车通行 | 13 |
| 禁止机动车通行 | 14 |
| 禁止非机动车通行 | 15 |
| 禁止鸣喇叭 | 16 |
| 立交直行和转弯行驶 | 17 |
| 限制速度40公里每小时 | 18 |
| 限速30公里每小时 | 19 |
| 鸣喇叭 | 20 |
| 其他 | 0 |
{
"system": {
"dataset": "myData", #数据集
"batch_size": 32, #batch_size
"sampling_function": "cornernet_saccade", #数据增强策略
"train_split": "train", #训练集
"val_split": "test", #验证集
"learning_rate": 0.00025, #初始学习率
"decay_rate": 10, #学习率衰减因子
"val_iter": 100, #每迭代val_iter计算一次val loss
"opt_algo": "adam", #优化器
"prefetch_size": 5, #队列预取数据量
"max_iter": 100000, #训练迭代的总次数
"stepsize": 200, #训练时每迭代stepsize次学习率衰减为原来的1/decay_rate
"snapshot": 5000, #训练每迭代snapshot次保存一次模型参数
"chunk_sizes": [
32
] #每块GPU上处理的图片数，其和等于batch_size
},
"db": {
"rand_scale_min": 0.5, #随机裁减比例[0.5,0.6，0.7，...,1.1]
"rand_scale_max": 1.1,
"rand_scale_step": 0.1,
"rand_scales": null,
"rand_full_crop": true, #随机裁剪
"gaussian_bump": true, #是否使用二维高斯给出惩罚减少量
"gaussian_iou": 0.5, #高斯半径的大小根据object尺寸得到,bounding box和gt box至少0.5IoU
"min_scale": 16,
"view_sizes": [],
"height_mult": 31,
"width_mult": 31,
"input_size": [
255,
255
], #网络输入图片的size
"output_sizes": [
[
64,
64
]
], #网络输出图片的size
"att_max_crops": 30, #和attention map相关的参数设置
"att_scales": [
[
1,
2,
4
]
],
"att_thresholds": [
0.3
], #概率>0.3的location被筛出来
"top_k": 12, #maximum number of crops to process
"num_dets": 12,
"categories": 10, #类别数
"ae_threshold": 0.3, #测试时，仅处理attention maps上 score > as_threshold=0.3 的locations
"nms_threshold": 0.5, #nms的阈值
"max_per_image": 100 #maximum number of objects to predict on a single image
}
}
import torch
import torch.nn as nn
from .py_utils import TopPool, BottomPool, LeftPool, RightPool
from .py_utils.utils import convolution, residual, corner_pool
from .py_utils.losses import CornerNet_Saccade_Loss
from .py_utils.modules import saccade_net, saccade_module, saccade
def make_pool_layer(dim):
return nn.Sequential()
def make_hg_layer(inp_dim, out_dim, modules):
layers = [residual(inp_dim, out_dim, stride=2)]
layers += [residual(out_dim, out_dim) for _ in range(1, modules)]
return nn.Sequential(*layers)
class model(saccade_net):
def _pred_mod(self, dim):
return nn.Sequential(
convolution(3, 256, 256, with_bn=False),
nn.Conv2d(256, dim, (1, 1))
)
def _merge_mod(self):
return nn.Sequential(
nn.Conv2d(256, 256, (1, 1), bias=False),
nn.BatchNorm2d(256)
)
def __init__(self):
stacks = 3
pre = nn.Sequential(
convolution(7, 3, 128, stride=2),
residual(128, 256, stride=2)
)
hg_mods = nn.ModuleList([
saccade_module(
3, [256, 384, 384, 512], [1, 1, 1, 1],
make_pool_layer=make_pool_layer,
make_hg_layer=make_hg_layer
) for _ in range(stacks)
])
cnvs = nn.ModuleList([convolution(3, 256, 256) for _ in range(stacks)])
inters = nn.ModuleList([residual(256, 256) for _ in range(stacks - 1)])
cnvs_ = nn.ModuleList([self._merge_mod() for _ in range(stacks - 1)])
inters_ = nn.ModuleList([self._merge_mod() for _ in range(stacks - 1)])
att_mods = nn.ModuleList([
nn.ModuleList([
nn.Sequential(
convolution(3, 384, 256, with_bn=False),
nn.Conv2d(256, 1, (1, 1))
),
nn.Sequential(
convolution(3, 384, 256, with_bn=False),
nn.Conv2d(256, 1, (1, 1))
),
nn.Sequential(
convolution(3, 256, 256, with_bn=False),
nn.Conv2d(256, 1, (1, 1))
)
]) for _ in range(stacks)
])
for att_mod in att_mods:
for att in att_mod:
torch.nn.init.constant_(att[-1].bias, -2.19)
hgs = saccade(pre, hg_mods, cnvs, inters, cnvs_, inters_)
tl_modules = nn.ModuleList([corner_pool(256, TopPool, LeftPool) for _ in range(stacks)])
br_modules = nn.ModuleList([corner_pool(256, BottomPool, RightPool) for _ in range(stacks)])
#这路需要修改为自己的类别数，我们这里是21个类别！！！
tl_heats = nn.ModuleList([self._pred_mod(21) for _ in range(stacks)])
br_heats = nn.ModuleList([self._pred_mod(21) for _ in range(stacks)])
for tl_heat, br_heat in zip(tl_heats, br_heats):
torch.nn.init.constant_(tl_heat[-1].bias, -2.19)
torch.nn.init.constant_(br_heat[-1].bias, -2.19)
tl_tags = nn.ModuleList([self._pred_mod(1) for _ in range(stacks)])
br_tags = nn.ModuleList([self._pred_mod(1) for _ in range(stacks)])
tl_offs = nn.ModuleList([self._pred_mod(2) for _ in range(stacks)])
br_offs = nn.ModuleList([self._pred_mod(2) for _ in range(stacks)])
super(model, self).__init__(
hgs, tl_modules, br_modules, tl_heats, br_heats,
tl_tags, br_tags, tl_offs, br_offs, att_mods
)
self.loss = CornerNet_Saccade_Loss(pull_weight=1e-1, push_weight=1e-1)