Only Truth Matters
0%

Posted on Edited on In , ,

1. 诗歌

我最早接触诗应该是在很小的时候，还在府庙的湖师附小。起因大概是语文老师有一次布置了诗歌的作业，让我们写诗，然后在班级里挑选了几个同学，参加什么诗歌的讨论组，小灶班？印象特别深的是参加这个讨论组应该是在周五，因为想到这件事，我仍能回忆起当时的开心，要不就是周五，要不就是能翘掉课，反正让人欣喜。

我已经无法想起我为什么会被选中，讨论组又究竟讨论了什么，只记得有一次讨论组布置了作业，我忘了写，于是在讨论之前几分钟随手写了几句，最后还被老师表扬了。大概是这样的：

秋天就像这片纸，

钻进了火里，

什么都没留下，

只留下了一缕烟尘

因为当时时间紧迫，我甚至没带纸，可能是找同学的本子撕了一小片纸写下来（这样的剧情在我的学生生涯中发生了许多许多次）

2. 审视

Ref

诗歌，或者说文字，也是人理解世界的一种方式。虽然我习惯用数字和物理去理解世界，但我相信，不同的理解方式能带给我新的认知

哲学，在希腊语中是爱智慧；诗，在希腊语中是被创造之物

诗歌的独特之处，诗歌易于分享，带有想象。

Read more »

Posted on Edited on In , , ,

1. Intro

Why this project? I wanna build a robot from scratch to fill the gap between theory and reality.

To build the robot, usually need following knowledge base

  • Mechanics, the build the body of robot
  • Electronic, the power of the robot
  • Computer, the brain of the robot

2. Goal

For the robot, I would like to focus on the mobile robot platform

Two level goals:

  1. Basic level, the robot can move, and use the lidar as a SLAM sensor, can track something using camera
  2. More precise control, going from A to B, the error should be less than 1cm

Consider a layer platform

  1. Environment sensing, different sensor to be used, a SBC is used. LiDAR, video camera, speaker, microphone
  2. Intelligence, communication, SBC, control other part
  3. Navigation, get the high-level command and go to desired position
Read more »

Posted on Edited on In , , ,

1. Intro

A model related to language, model behind GPT, BERT, T5

2. What

A general model

  • CNN for vision
  • RNN for language, sequential
  • hard to train
  • always forget

Initially, trained for translation

  • good for training
  • huge dataset

3. How it work

  1. Positional Encoding

add a index for the word, easier to train

Read more »

Posted on Edited on In , , ,

1. Alpha Matting

Goal: accurate matting for fine details

Application: background replacement, synthetic bokeh

Idea: not binary, but estimate a transparency value

How: Train a CNN

  • input: RGB image and coarse alpha matte
  • MobileNetV3 and shallow encoder-decoder
  • A studio to make high-quality dataset

2. Omnimattes

Goal: produce mask not only for objects but also their effects (shadow, smoke, reflection..)

Application: image and video editing

Idea: create different layer for each object and also the background

Read more »

Posted on Edited on In , , ,

1. Why

Method: test the single motor first and then deploy it to real robot

Any problem can be solved using a layered abstraction

2. Motor Types

AC Motor, taken from AC, application in industrial

  • Higher power, higher torque

Brushed and Brushless DC

  • Brushed
  • Brushless: smooth, no noise, efficient, high speed

Geared-: Speed reduction

Stepper/Servo: highest precision

Read more »

Posted on Edited on In , ,

1. Intro

Common health problem

2. Eye problem & Migraines

headache, dizziness, computer vision syndrome

Suggestion:

  1. 20-20-20
  2. Good lighting
  3. Blue light blocking glasses
  4. Adjust monitor

3. Back, neck and shoulder

pain, stiff, spinal

Suggestion:

  1. sit/stand
  2. stretch break
  3. good posture/chair
Read more »

Posted on Edited on In , ,

1. Lean Startup

Goal: create something customer would buy, as quick as possible

No idea is that great

The main thing is execution.

Takeaway

  1. build-measure-learn feedback loop
  2. everything is a grand experiment
  3. observe, don’t ask
  4. MVP
  5. from easy to real
  6. Three engine of growth
  7. sticky, long term customer
  8. viral, new customer should invite others
  9. paid
  10. Pivot?

2. Fastlane to financial freedom

From current to freedom

  1. use the business and investment, time is the highest value, create passive income, and chase dream
  2. wealth=profit+asset value, not bound by time
  3. five bussiness
  4. control
  5. entry
  6. need
  7. time
  8. scale: quantity or quality

3. Software Developer

Read more »