BiTraP：Bi-directional Pedestrian Trajectory Prediction with Multi-modal Goal Estimation


import torch
import torch.nn as nn
import torch.nn.functional as F# 先验网络
class Prior(nn.Module):def __init__(self,input_size=256,output_size=64):super(Prior,self).__init__()self.input_size = input_size # 输入大小self.output_size = output_size # 输出大小self.fc1 = nn.Linear(input_size,128)self.fc2 = nn.Linear(128,64)self.fc3 = nn.Linear(64,output_size)def forward(self,x):# x：[bs,256]x = F.relu(self.fc1(x))x = F.relu(self.fc2(x))x = self.fc3(x)return x# 后验网络
class Recognition(nn.Module):def __init__(self,input_size=512,output_size=64):super(Recognition,self).__init__()self.input_size = input_size # 输入大小self.output_size = output_size # 输出大小self.fc1 = nn.Linear(input_size,256)self.fc2 = nn.Linear(256,128)self.fc3 = nn.Linear(128,output_size)def forward(self,x):# x：[bs,512]x = F.relu(self.fc1(x))x = F.relu(self.fc2(x))x = self.fc3(x)return x# 目标网络
class Goal(nn.Module):def __init__(self,input_size=256+32,output_size=2):super(Goal,self).__init__()self.input_size = input_size # 输入大小self.output_size = output_size # 输出大小self.fc1 = nn.Linear(input_size,128)self.fc2 = nn.Linear(128,64)self.fc3 = nn.Linear(64,output_size)def forward(self,x):# x：[bs,256+32]x = F.relu(self.fc1(x))x = F.relu(self.fc2(x))x = self.fc3(x)return xclass BiTrap(nn.Module):def __init__(self,embedding_size=64,input_size=2,output_size=2,gru_size=256,latent_size=32,gru_input_size=64,num_layer=1,pre_len=12):super(BiTrap,self).__init__()self.input_size = input_size # 输入大小self.output_size = output_size # 输出大小self.embedding_size = embedding_size # 空间编码self.gru_size = gru_size # GRU隐藏层大小self.latent_size = latent_size # z的大小self.gru_input_size = gru_input_size # GRU输入大小self.num_layer = num_layer # GRU层数self.pre_len = pre_len # 预测长度# 观测轨迹升维self.fcx = nn.Linear(input_size,embedding_size)# 预测轨迹升维self.fcy = nn.Linear(input_size,embedding_size)# 观测轨迹GRUself.grux = nn.GRU(embedding_size,gru_size)# 预测轨迹GRUself.gruy = nn.GRU(embedding_size,gru_size)# 先验网络 后验网络 目标网络self.prior = Prior(gru_size,2*latent_size)self.recognition = Recognition(gru_size*2,2*latent_size)self.goal = Goal(gru_size+latent_size,output_size)# 状态转换网络self.fcf = nn.Linear(gru_size+latent_size,gru_input_size)self.fc2 = nn.Linear(gru_size+latent_size,gru_size)self.fc3 = nn.Linear(gru_size,gru_input_size)self.fc4 = nn.Linear(gru_size+latent_size,gru_size)self.fc5 = nn.Linear(output_size,gru_input_size)# GRUcellself.forward_gru = nn.GRUCell(gru_input_size,gru_size)self.backward_gru = nn.GRUCell(gru_input_size,gru_size)# 最后的输出self.fc6 = nn.Linear(2*gru_size,output_size)def forward(self,x,mode='train',y=None):n,s = x.shape[-1],x.shape[2]# [bs,c,seq_len,n]->[seq_len,n,c]c=2x = x.squeeze(0).permute(1,2,0)# [seq_len*n,c]x = x.reshape(-1,x.shape[-1])# [seq_len*n,embedding]x = self.fcx(x)# [seq_len,n,embedding]x = x.reshape(s,n,-1)x_gru_h = torch.randn(self.num_layer,n,self.gru_size).cuda()_,h = self.grux(x,x_gru_h)# [n,embedding]x = h.squeeze(0)# 重参数化e = torch.randn(n,self.latent_size).cuda()# 求得先验分布p = self.prior(x)if(mode=='train'):n,s = y.shape[-1],y.shape[2]# [bs,c,seq_len,n]->[seq_len,n,c]c=2y = y.squeeze(0).permute(1,2,0)# [seq_len*n,c]y = y.reshape(-1,y.shape[-1])# [seq_len*n,embedding]y = self.fcy(y)# [seq_len,n,embedding]y = y.reshape(s,n,-1)y_gru_h = torch.randn(self.num_layer,n,self.gru_size).cuda()_,y = self.gruy(y,y_gru_h)# [n,embedding]y = y.squeeze(0)# [n,2*embeddin]y = torch.cat((x,y),1)# 求得后验分布q = self.recognition(y)z = q[:,0:self.latent_size]+q[:,self.latent_size:]*eelse:z = p[:,0:self.latent_size]+p[:,self.latent_size:]*e# [n,gru_size+latent_size]x = torch.cat((h.squeeze(0),z),1)# 求Goal [n,2]g = self.goal(x)forward_gru_h = self.fc2(x)f = self.fcf(x) # 前项输入backward_gru_h = self.fc4(x)b = self.fc5(g) # 后项输入# 计算前向forward_output = []for i in range(self.pre_len):forward_gru_h = self.forward_gru(f,forward_gru_h)forward_output.append(forward_gru_h)f = self.fc3(forward_gru_h)# 计算后向backward_output = []for i in range(self.pre_len-1,-1,-1):backward_gru_h = self.backward_gru(b,forward_gru_h)temp = torch.cat((forward_output[i],backward_gru_h),1)output = self.fc6(temp)backward_output.append(output)b = self.fc5(output)if(mode=='train'):return p,q,g,torch.stack(backward_output,0).unsqueeze(0)else:return torch.stack(backward_output,0).unsqueeze(0)"""
x = torch.randn(1,2,8,24)
y = torch.randn(1,2,12,24)
prior = BiTrap(embedding_size=64,input_size=2,output_size=2,gru_size=256,latent_size=32,gru_input_size=64)
p,q,g,b = prior(x,'train',y)
print(p.shape)
print(q.shape)
print(g.shape)
print(b.shape)b = prior(x,'test')
print(b.shape)
"""

BiTraP：Bi-directional Pedestrian Trajectory Prediction with Multi-modal Goal Estimation相关推荐

行人轨迹论文阅读SSAGCN: Social Soft Attention Graph Convolution Network for Pedestrian Trajectory Prediction
SSAGCN: Social Soft Attention Graph Convolution Network for Pedestrian Trajectory Prediction SSAGCN: ...
论文翻译 SGCN:Sparse Graph Convolution Network for Pedestrian Trajectory Prediction 用于行人轨迹预测的稀疏图卷积网络
SGCN:Sparse Graph Convolution Network for Pedestrian Trajectory Prediction 用于行人轨迹预测的稀疏图卷积网络行人轨迹预测是自 ...
文献翻译：Social LSTM: Human Trajectory Prediction in Crowded Spaces
这是我阅读的有关轨迹预测的第一篇文献,其内容和使用的模型相对简单,是比较适合的入门篇,我在此把原文翻译分享出来,便于大家交流学习. 这里写目录标题 Abstract ...
【ECCV2020】Spatio-Temporal Graph Transformer Networks for Pedestrian Trajectory Prediction
[ECCV2020]用于行人轨迹预测的时空图 Transformer 网络摘要了解人群运动动力学对于现实世界的应用至关重要,例如监控系统和自动驾驶.这是具有挑战性的,因为它需要对具有社会意识的人群 ...
评估行人行动预测的基准——Benchmark for Evaluating Pedestrian Action Prediction
评估行人行动预测的基准--Benchmark for Evaluating Pedestrian Action Prediction Date of Conference: 3-8 Jan. 2021 ...
[论文阅读]用于车辆轨迹预测的卷积社交池Convolutional Social Pooling for Vehicle Trajectory Prediction
文章目录一.摘要二.介绍三.相关研究 3.1 基于机动的模型 3.2 交互感知模型 3.3 运动预测的递归网络四.问题制定 4.1 参照系 4.2 输入输出 4.3 概率运动预测 4.4 操作 ...
顶会论文笔记：联邦学习——ATPFL: Automatic Trajectory Prediction Model Design under Federated Learning Framework
ATPFL: Automatic Trajectory Prediction Model Design under Federated Learning Framework 文章目录 ATPFL: A ...
文献阅读笔记：EvolveGraph: Multi-Agent Trajectory Prediction with Dynamic Relational Reasoning
文献阅读笔记摘要 1 引言 2 相关工作 3 Problem formulation 4 EvolveGraph 5 Experiments 6 结论 EvolveGraph: Multi-Agen ...
Social LSTM: Human Trajectory Prediction in Crowded Spaces 论文翻译
摘要行人可沿不同的轨道行走,以避开障碍物及方便其他行人.在这样的场景中行驶的任何自动驾驶车辆都应该能够预见行人未来的位置,并相应地调整其路径以避免碰撞.轨迹预测问题可以看作是一个序列生成任务,我们感 ...

BiTraP：Bi-directional Pedestrian Trajectory Prediction with Multi-modal Goal Estimation

BiTraP：Bi-directional Pedestrian Trajectory Prediction with Multi-modal Goal Estimation相关推荐

最新文章

热门文章