- Jetpack 4.3
- SD card version is r32.3.1
- Other lib versions (e.g. cuda, cuDNN, tensorrt...) showed in here
- Japanese languages
- ROS and so on
- Flash Jetbot images to SD card
""" | |
Python: 3.9.5 | |
torch: 1.12.1+cu113 | |
pytorch3d: 0.7.1 | |
""" | |
import torch | |
from torch import nn | |
from pytorch3d.transforms.rotation_conversions import quaternion_to_matrix, matrix_to_quaternion | |
def broadcast_1d_to_2dtile(x, h_and_w_list, name=None): | |
""" | |
Convert (Batch, dim) to the image like tensor, as (Batch, h, w, dim) | |
That function is used in "Neural Representation and rendering (S. M. Ali Eslami+ 2018 Science)" | |
to feed CNN "v" vector. | |
e.g.) x=[[1,2,3]] and batch_size=1 h_and_w_list=[2,2] | |
=> [[[1,1],[1,1]],[[2,2],[2,2]],[[3,3],[3,3]]] | |
:param x: A tensor, (Batch, dim) | |
:param h_and_w_list:A list of output image, shape is [h, w] |
def tf_softargmax(x, batch ,min=0., max=1.): | |
""" | |
"soft-argmax" is a differentiable version of "argmax" function [Kendall et al. 2017]. | |
URL: https://arxiv.org/abs/1703.04309 | |
First computes the softmax over the spatial extent of each channel of a convolutional feature map. | |
Then computes the expected 2D position of the points of maximal activation for each channel, | |
resulting in a set of feature keypoints (e.g. [[E[x1], E[y1]], [E[x2], E[y2]], ... ,E[xN], E[yN]]) | |
in each channel and batch. |
\begin{document}
\pagestyle{empty}
\tableofcontents%目次のページ番号を削除したい
\clearpage%%%これをいれるのが重要.
%いれないと目次の最終ページがどうなるか見てましょう.
\pagestyle{plain}%%jreportのデフォルトのpagestyle
\setcounter{page}{1}%%これで,1章1ページ目が-1-に