影片

\(t\)

CNN

CNN

CNN

CNN

圖片特徵

圖片特徵

圖片特徵

圖片特徵

RNN

音效特徵

音效特徵

音效特徵

音效特徵

音效特徵

音效特徵

音效特徵

音效特徵

音效特徵

音效特徵

音效特徵

音效特徵

音效特徵

資料庫片段

資料庫片段

資料庫片段

資料庫片段

\rbrace
\rbrace
\rbrace
\rbrace
\rbrace

資料庫片段

下載圖片

圖片前處理

蒐集訓練資料

建構類神經網路

訓練

討論模型表現

訓練模型

擷取畫面

標記畫面

根據標記配音效

影片配音效

調整模型

根據標記配音效

討論成效

成果

1.原始網路

2.去除決策網路

3.訓練決策網路

4.微調

ImageNet

extraction

decision

prediction1

20 classes

extraction

decision

prediction2

transfer

(retrain)

ResNet152V2

(1000 classes)

new model

frames

CNN

fire

labels

ResNet

 ALGORITHM

fire

ResNet

-label-

ResNet

-label-

ResNet

ambulance

ResNet

ambulance

ResNet

ambulance

fire

soundeffect

audios

\text{Video}\begin{cases}\begin{array}{llll} \text{Img}_1 & \xrightarrow{\text{ResNet}} & \text{Class}_1\\ \text{Img}_2 & \xrightarrow{\text{ResNet}} & \text{Class}_2\\ \vdots & & \vdots\\ \text{Img}_{t-1} & \xrightarrow{\text{ResNet}} & \text{Class}_{t-1}\\ \text{Img}_t & \xrightarrow{\text{ResNet}} & \text{Class}_t\\ \end{array}\end{cases}
\text{Video}\begin{cases}\begin{array}{llll} \text{Img}_1 & \xrightarrow{\text{ResNet}} & \text{Class}_1\\ \text{Img}_2 & \xrightarrow{\text{ResNet}} & \text{Class}_2\\ \vdots & & \vdots\\ \text{Img}_{t-1} & \xrightarrow{\text{ResNet}} & \text{Class}_{t-1}\\ \text{Img}_t & \xrightarrow{\text{ResNet}} & \text{Class}_t\\ \end{array}\end{cases}

\(\text{Class}_t:\)

t=1\qquad A
\vdots

\(\text{Sound}_A\)

\(\text{Sound}_B\)

t=2\qquad A
t=3\qquad A
t=4\qquad A
t=5\qquad B
t=6\qquad B

舊畫面人物

新畫面人物

cost (distance)

vanish cost

消失節點

vanish cost

cost (distance)

舊畫面人物

消失節點

新畫面人物

\(e^-\)

\(e^-\)

\(e^-\)

\(\gamma\)

\(\gamma\)

\(Q_e\)

\(Q_e\)

\(\gamma\)

\(e^+\)

\(e^+\)

\(e^-\)

\(e^-\)

\(Q_e\)

\(Q_e\)

\(W^+\)

\(\nu_e\)

\(e^-\)

\(e^-\)

\(\nu_e\)

\(a\)

\(\sqrt{3}a\)

arr[0]
arr[1]
...
arr[ptr-1]
arr[ptr]
arr[ptr+1]

\(\uparrow\)

ptr

 Stack data

arr[head]
arr[head+1]
...
...
arr[tail-1]
arr[tail]

\(\uparrow\)

tail

 Queue data

arr[head-1]
arr[tail+1]

\(\uparrow\)

head
prev
next
data
prev
next
data
prev
next
data
prev
next
data
prev
next
data
NULL
NULL
Node
Node
Node
Node
Node

Linked List data

prev
next
data
prev
next
data
prev
next
data
NULL
NULL
Node
Node
Node
Node
Node
prev
next
prev
next

\(\uparrow\)

tail

\(\uparrow\)

head

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

41

34

62

0

45

78

5

58

69

99

0

5

24

58

34

41

78

67

62

69

64

45

0

5

24

58

34

41

78

67

62

69

64

45

17

5

24

58

34

41

78

67

62

69

64

45

before

after

\((t, p)\)

\((t', p')\)

\(T\)

before

after

\((t, p)\)

\((t', p')\)

\(T\)

\(T'\)

before

after

\((t, p)\)

\((t', p')\)

\(T\)

\(T'\)

before

after

\((t', p')\)

\((t, p)\)

\(T\)

before

after

\((t, p)\)

\((t', p')\)

\(T\)

\(\hat{T}\)

\((\hat{t}, \hat{p})\)

1,3

1,2

1,1

2,3

2,2

2,1

3,3

3,2

3,1

4,1

1,2

1,1

2,3

2,2

2,1

3,3

3,2

4,1

1,2

1,1

2,3

2,2

2,1

3,3

3,2

4,1

  • Wifi名稱：CSIE_guest

  • EAP Method：PEAP

  • Phase 2 authentication：MSCHAPV2

  • 身分 Identity：ioic2023

  • 匿名身分 Anonymous identity：（免填）

  • 密碼 Password：CNS5DS2D

\(D\)

\(A'\)

\(C\)

\(D\)

\(B'\)

\(C\)

\(D\)

\(A\)

\(C'\)

\(D\)

\(A'\)

\(B'\)

\(C'\)

\(C\)

\(A\)

A

BC

0

1

00

01

11

10

A

BC

0

1

00

01

11

10

1

1

1

X

X

X

0

0

AB

CD

00

01

11

10

00

01

11

10

AB

CD

00

01

11

10

00

01

11

10

0

0

0

0

0

0

0

X

X

1

1

1

1

1

1

1

AB

CD

00

01

11

10

00

01

11

10

AB

CD

00

01

11

10

00

01

11

10

0

0

0

0

0

0

0

X

X

1

1

1

1

1

1

1

AB

CD

00

01

11

10

00

01

11

10

AB

CD

00

01

11

10

00

01

11

10

1

1

1

0

0

0

X

X

1

0

1

1

1

1

1

0

AB

CD

00

01

11

10

00

01

11

10

AB

CD

00

01

11

10

00

01

11

10

0

0

0

0

0

0

0

1

0

1

1

1

1

0

0

0

AB

CD

00

01

11

10

00

01

11

10

0

1

1

1

1

0

0

1

1

1

1

0

1

0

0

1

AB

CD

00

01

11

10

00

01

11

10

0

1

1

1

1

0

0

1

1

1

1

0

1

0

0

1

S_0
S_1
S_2
S_3
\left\{A, C\right\}

\(C_1\)

\left\{A\right\}

\(C_2\)

\left\{B, C\right\}

\(C_3\)

\left\{B, C\right\}

\(C_4\)

\left\{A, B\right\}

\(C_5\)

\left\{C\right\}

\(C_1\)

\left\{A\right\}

\(C_2\)

\left\{B, C\right\}

\(C_3\)

\left\{B, C\right\}

\(C_4\)

\left\{A, B\right\}

\(C_5\)

Alice

pizza

quesadillas

ramen

sushi

Chris

pizza

quesadillas

ramen

Bob

pizza

ramen

=

David

pizza

ramen

sushi

\neq
\neq
\neq
\neq

17

1

13

6

9

9

\(\alpha, v, \beta\)

\(9, 9, \infty\)

\(-\infty, 9, 9\)

\(17, 17, 9\)

\(9, 9, \infty\)

\(-\infty, 17, 9\)

\(9, 7, 7\)

\(9, 7, \infty\)

\(\alpha, v, \beta\)

\(9, 9, \infty\)

\(-\infty, 6, 6\)

\(6, 6, \infty\)

\(13, 13, 6\)

\(-\infty, 13, 6\)

\(7, 7, 6\)

\(17, 17, \infty\)

\(6, 9, 9\)

\(9, 9, 17\)

\(15, 17, 17\)

\(15, 11, 11\)

\(6, 15, 15\)

Max

Expected

Expected

Expected

Expected

\(\cdots\)

fold

call

raise 2

raise 1024

Min

Min

Min

Min

Min

Min

Min

Min

Min

Min

Min

Min

Min

Min

Min

Min

Min

Min

Min

Min

Min

Min

Min

Min

inode

address0

address1

address2

\(\vdots\)

address10

address11

address12

data

data

address0

address1

\(\vdots\)

address4

address255

\(\vdots\)

address0

\(\vdots\)

address94

address255

\(\vdots\)

data

  66666
= 11 + 66655
= 11 + 65536 + 1119
= 11 + 65536 + 1024 + 95

address0~255

address0~255

pretrained

transformer

Scalar

Token

Textual

sentence

representation

\(W_{768\times3}\)

ReLU

average

scroing

average

dancibility

fill missings

normalization

 

Deep Learning Model

Dancibility

Raw Data

Trainables

Model Inputs

normalized

features

系上資訊站

工作坊

讀document會

訪談

 新生手冊

課程完成度須達 100%

解題進度須達標準

須蓋章完成

追蹤社團 IG

username

轉發貼文

client

dash.mpd

server

setup

playing

init-stream1.m4s
chunk-stream1-00001.m4s
chunk-stream1-00002.m4s
chunk-stream1-00003.m4s
chunk-stream1-00004.m4s

long seek / back seek / change resolution

0

7

12

15

20

25

32

Hazard

IF

ID

EX

MEM

WB

Fwd

ID

IF

HD

EX

EX

Messenger

Instagram

Line

Chrome

+

=

Default Weekly Schedule

Original

Result

Made with Slides.com