profile
์ฝ” ๋”ฉ ์กฐ ์•„ ๐Ÿ‘ฉ๐Ÿปโ€๐Ÿ’ป ๋ฐฐ์›Œ์„œ ๋‚จ์ฃผ์ž ๐Ÿ’
post-thumbnail

[ML] ๋จธ์‹ ๋Ÿฌ๋‹ ๊ธฐ์ดˆ

์ธ๊ฐ„์˜ ํ•™์Šต๋Šฅ๋ ฅ, ์ถ”๋ก  ๋Šฅ๋ ฅ๋“ฑ์„ ์ปดํ“จํ„ฐ๋ฅผ ํ†ตํ•ด ๊ตฌํ˜„ํ•˜๋Š” ํฌ๊ด„์  ๊ฐœ๋…Strong AI ์™€ week AI ๋กœ ๊ตฌ๋ถ„Data ๋ฅผ ์ด์šฉํ•˜์—ฌ ๋ฐ์ดํ„ฐ ํŠน์„ฑ๊ณผ ํŒจํ„ด์„ ํ•™์Šตํ•˜์—ฌ, ๊ทธ ๊ฒฐ๊ณผ๋ฅผ ๋ฐ”ํƒ•์œผ๋กœ ํŠน์ • ๋ฐ์ดํ„ฐ์— ๋Œ€ํ•œ ๋ฏธ๋ž˜ ๊ฒฐ๊ณผ(๊ฐ’, ๋ถ„ํฌ)๋ฅผ ์˜ˆ์ธกLearn from data๋จธ์‹ ๋Ÿฌ๋‹์˜

2์ผ ์ „
ยท
0๊ฐœ์˜ ๋Œ“๊ธ€

python error: PIL.Image.DecompressionBombError

PIL.Image.DecompressionBombError: Image size (190476000 pixels) exceeds limit of 178956970 pixels, could be decompression bomb DOS attack.์œ„์™€ ๊ฐ™์€ ์—๋Ÿฌ๊ฐ€ ๋ฐœ์ƒ

2021๋…„ 6์›” 24์ผ
ยท
0๊ฐœ์˜ ๋Œ“๊ธ€

21.05.31

-

2021๋…„ 5์›” 31์ผ
ยท
2๊ฐœ์˜ ๋Œ“๊ธ€
post-thumbnail

TOPCIT 15ํšŒ ์ •๊ธฐํ‰๊ฐ€ ํ›„๊ธฐ

์ง€๊ธˆ๊นŒ์ง€ 3๋ฒˆ์งธ ๋ณด๋Š” ํƒ‘์‹ฏ์ธ๋ฐ ๋‹จ ํ•œ ๋ฒˆ๋„ ๊ณต๋ถ€๋ฅผ ํ•˜๊ณ  ๋ณด๋Ÿฌ ๊ฐ„ ์ ์ด ์—†์–ด์„œ ํ›„๊ธฐ๋ผ๊ณ  ํ•˜๊ธฐ๋„ ์ข€ ๊ทธ๋ ‡๊ณ , ์–ด๋–ค ๋ฌธ์ œ๊ฐ€ ์žˆ์—ˆ๋Š”์ง€ ๋˜์งš์–ด ๋ณด๋Š” ๊ธ€ ์ •๋„๋กœ ํ•ด๋‘ฌ์•ผ๊ฒ ๋‹ค. ๊ณต๋ถ€๋„ ํ•˜๋‚˜๋„ ์•ˆํ•˜๊ณ  ๊ฐ”์œผ๋ฉด์„œ ์›ฌ ํ›„๊ธฐ๋ƒ๊ตฌ์š”? ๋‹ค์Œ ํ•˜๋ฐ˜๊ธฐ ํŠน๋ณ„ํ‰๊ฐ€ ๋•Œ๋Š” ์—ด์‹ฌํžˆ ๋ณผ ์˜ˆ์ •์ด๊ธฐ ๋•Œ๋ฌธ์— ๋‚ด๊ฐ€

2021๋…„ 5์›” 22์ผ
ยท
0๊ฐœ์˜ ๋Œ“๊ธ€

Mixture of Logistic Distributions์ด๋ž€

Pixel CNN++ ๋…ผ๋ฌธ์—์„œ softmax๋ง๊ณ  discretized logestic mixture likkelihood๋ฅผ ์‚ฌ์šฉํ–ˆ๋‹ค๋Š”๋ฐ, ์ด๊ฒŒ ๋ญ”์ง€ ๋ชฐ๋ผ์„œ ์ฐพ์•„ ๋ณด์•˜๋‹ค.

2021๋…„ 4์›” 27์ผ
ยท
0๊ฐœ์˜ ๋Œ“๊ธ€

polling๊ณผ interrupt์˜ ์ฐจ์ด์ 

๋ณด๋“œ๋ฅผ ์“ธ ๋•Œ ์‚ฌ์šฉ์ž์˜ ์ž…๋ ฅ์„ polling/interrupt ๋ฐฉ์‹์— ์˜ํ•ด ๊ฐ์ง€ํ•  ์ˆ˜ ์žˆ๋‹ค.2๊ฐ€์ง€ ๋ฐฉ์‹์œผ๋กœ ๊ตฌํ˜„ํ•˜๊ธฐ์— ์•ž์„œ ๊ฐœ๋…์„ ์ •์˜ํ•ด๋ณด์ž.polling์€ ํ•˜๋“œ์›จ์–ด์˜ ๋ณ€ํ™”๋ฅผ ์ง€์†์ ์œผ๋กœ ์ฝ์–ด ๋“ค์ด๋ฉฐ ์ด๋ฒคํŠธ์˜ ์ˆ˜ํ–‰ ์—ฌ๋ถ€๋ฅผ ์ฃผ๊ธฐ์ ์œผ๋กœ ๊ฒ€์‚ฌํ•˜์—ฌ ํ•ด๋‹น ์‹ ํ˜ธ๋ฅผ ๋ฐ›์•˜์„๋•Œ ์ด๋ฒคํŠธ๋ฅผ

2021๋…„ 4์›” 26์ผ
ยท
2๊ฐœ์˜ ๋Œ“๊ธ€

Auto-Regressive model study

AR ๋ชจ๋ธ์— ๋Œ€ํ•ด ๊ณต๋ถ€๋ฅผ ํ•˜์ž ๊ณต!๋ถ€!Pixel RNN: Row LSTM, Diagonal BiLSTM, Pixel CNNPixel CNN decoders: Gated PixelCNNAuto-Regressive model์€ model์˜ Output์„ ๋‹ค์‹œ input์œผ๋กœ

2021๋…„ 4์›” 26์ผ
ยท
0๊ฐœ์˜ ๋Œ“๊ธ€

[OODP] RMI

๋ถ„์‚ฐ๋˜์–ด ์กด์žฌํ•˜๋Š” ๊ฐ์ฒด ๊ฐ„์˜ ๋ฉ”์‹œ์ง€ ์ „์†ก์„ ๊ฐ€๋Šฅํ•˜๊ฒŒ ํ•จ (๋ฐ์ดํ„ฐ๋ฒ ์ด์Šค, ์• ํ”Œ๋ฆฌ์ผ€์ž‡์…˜, ํด๋ผ์ด์–ธํŠธ ๋ชจ๋‘ ๊ฐ์ฒด๋ฅผ ํ†ตํ•ด ๊ตฌํ˜„ํ•˜๊ณ  ์ €์žฅํ•˜๊ฒŒ ๋œ๋‹ค. ์ด๋ ‡๊ฒŒ ๋ถ„์‚ฐ๋˜์–ด ์กด์žฌํ•˜๊ณ  ์žˆ๋Š” ๊ฐ์ฒด๋“ค ๊ฐ„์˜ ๋ฉ”์‹œ์ง€ ์ „์†ก์˜ ํ•„์š”์„ฑ์ด ์ƒ๊ฒผ๊ณ , ์ด๋ฅผ ๊ฐ€๋Šฅํ•˜๊ฒŒ ํ•˜๋Š” ๊ฒƒ์ด RMI ํ”„๋กœํ† ์ฝœ์ด๋‹ค.RMI์ž์ฒด

2021๋…„ 4์›” 12์ผ
ยท
0๊ฐœ์˜ ๋Œ“๊ธ€
post-thumbnail

Pixel Recurrent Neural Networks ๋…ผ๋ฌธ ๋ฆฌ๋ทฐ

๋” ์ด์ƒ ๋ฏธ๋ฃฐ ์ˆ˜ ์—†์–ด!! 0. Abstract Natural image์˜ distribution์„ ๋ชจ๋ธ๋ง ํ•˜๋Š” ๊ฒƒ์€ unsupervised learning์˜ ๋‘๋“œ๋Ÿฌ์ง„ ๋ฌธ์ œ์ด๋‹ค. ์ด๋Ÿฐ task์—๋Š” ํ‘œํ˜„๋ ฅ์ด ๋›ฐ์–ด๋‚˜๊ณ  ๋‹ค๋ฃจ๊ธฐ ์‰ฝ๊ณ  scalableํ•œ ์ด๋ฏธ์ง€ ๋ชจ๋ธ์ด ํ•„์š”ํ•˜

2021๋…„ 4์›” 12์ผ
ยท
1๊ฐœ์˜ ๋Œ“๊ธ€

21.03.27

์‹œ๊ฐ„์ด ๋„ˆ๋ฌด ๋นจ๋ฆฌ ํ๋ฅด๊ณ  ๋ฐ”์œ ํ•˜๋ฃจํ•˜๋ฃจ๋ฅผ ๋ณด๋‚ด์ง€๋งŒ ๋ฌด์—‡์„ ์–ด๋–ป๊ฒŒ ๊ณต๋ถ€ํ•˜๊ณ  ์žˆ๊ณ , ์ œ๋Œ€๋กœ ์‹ค๋ ฅ์„ ์Œ“์•„๊ฐ€๊ณ  ์žˆ๋Š”์ง€ ์˜๋ฌธ์ด ๋“ค์–ด์„œ ์‹œ์ž‘ํ•˜๋Š” TIL.์‹œ๊ฐ„ ๊ด€๋ฆฌ๋ฅผ ๋˜‘๋ฐ”๋กœ ํ•˜๊ณ , ์ฃผ์–ด์ง„ ์ผ๋“ค์„ ์ง‘์ค‘ํ•ด์„œ ๋นจ๋ฆฌ ๋๋‚ผ ๊ฒƒ.์ž˜ํ•ด์•ผ ํ•˜๋Š” ๊ฒƒ์ด ๋„ˆ๋ฌด ๋งŽ์€ 4ํ•™๋…„ ์ปด๊ณต. ํž˜๋“ค์ง€๋งŒ ์–ด์ฉ” ์ˆ˜ ์—†

2021๋…„ 3์›” 27์ผ
ยท
5๊ฐœ์˜ ๋Œ“๊ธ€
post-thumbnail

SROBB: Targeted Perceptual Loss for Single Image Super-Resolution

์ตœ๊ทผ percptual loss ๊ธฐ๋ฐ˜ super resolution ์—ฐ๊ตฌ๋“ค์€ ์„ฑ๋Šฅ ํ–ฅ์ƒ์„ ์ด๋ค„์™”๋‹ค. ์ด์™€ ๊ฐ™์€ objective function๋“ค์€ ๊ฑฐ์˜ ์‚ฌ์ง„๊ณผ ์œ ์‚ฌํ•œ ๊ฒฐ๊ณผ๋ฅผ ์ƒ์„ฑํ•œ๋‹ค. ํ•˜์ง€๋งŒ ์ด๋Š” ์ด๋ฏธ์ง€ ๋‚ด์˜ semantic information๋“ค์„ ๊ณ ๋ คํ•˜์ง€ ์•Š๊ณ ,

2021๋…„ 3์›” 23์ผ
ยท
0๊ฐœ์˜ ๋Œ“๊ธ€

Keil MDK-ARM v5 (ฮผVision5) ์„ค์น˜

์ž„๋ฒ ๋””๋“œ ์–ด์…ˆ๋ธ”๋ฆฌ ๊ณต๋ถ€๋ฅผ ํ•˜๊ธฐ ์œ„ํ•ด ํ•„์š”ํ•œ mdk-arm v5๋ฅผ ์„ค์น˜ํ•ด๋ณด๊ฒ ๋‹ค. ์ปดํ“จํ„ฐ๊ตฌ์กฐ์—์„œ MIPS ์ฝ”๋”ฉํ• ๋•Œ๋Š” QTspim ์ด๋ฆ„ ์ง„์งœ ์กธ๊ท€ํƒฑ์ธ๊ฑฐ ์‚ฌ์šฉํ–ˆ์—ˆ๋Š”๋ฐ, ์ด๊ฑด ์ฒ˜์Œ๋ณธ๋‹ค.. ๋ถ€์ง€๋Ÿฐํžˆ ์—ด๊ณต ๋š๋”ฑ๋š๋”ฑ! ์ž„๋ฒ ๋””๋“œ ๋ฟŒ์Š๋น ์ˆ‘! Keil MDK-ARM v5 ๋‹ค์šด๋กœ๋“œ ke

2021๋…„ 3์›” 8์ผ
ยท
0๊ฐœ์˜ ๋Œ“๊ธ€
post-thumbnail

๋”ฅ๋Ÿฌ๋‹ ํ•™์Šต์„ ์œ„ํ•œ GPU ์„œ๋ฒ„ ๊ตฌ์ถ•

๋ฐฉํ•™๋™์•ˆ ์ธํ„ด์œผ๋กœ ๊ทผ๋ฌดํ•œ ํšŒ์‚ฌ์—์„œ ๋”ฅ๋Ÿฌ๋‹ ํ•™์Šต์„ ์œ„ํ•ด RTX 3090 GPU๋ฅผ ์ฃผ๊ณ (์ง„์งœ ์ข‹์•„ 3090,, ๋‚˜ ํ˜ผ์ž ํ•™์Šต ๋งˆ๊ตฌ๋งˆ๊ตฌ>\_<), ์„œ๋ฒ„๋ฅผ ๊ตฌ์ถ•ํ•˜์—ฌ ์‚ฌ์šฉํ•˜๋ผ๊ณ  ํ•˜์…จ๋‹ค. ํ•™๊ต ์—ฐ๊ตฌ์‹ค์—์„œ๋Š” ์„œ๋ฒ„ ๊ด€๋ฆฌ๋ฅผ ๊ต์ˆ˜๋‹˜๊ณผ ๋Œ€ํ•™์›์ƒ ๋ถ„๋“ค๊ป˜์„œ ํ•ด์ฃผ์…จ๋Š”๋ฐ ์ฒ˜์Œ์œผ๋กœ ํ•ด๋ณด๋Š” ์ž‘

2021๋…„ 3์›” 6์ผ
ยท
0๊ฐœ์˜ ๋Œ“๊ธ€
post-thumbnail

DiffAugmentation ๋…ผ๋ฌธ ์ •๋ฆฌ

GAN์„ ํ•™์Šต์‹œํ‚ฌ ๋•Œ ์ œํ•œ๋œ ์–‘์˜ ํ›ˆ๋ จ ๋ฐ์ดํ„ฐ๊ฐ€ ์‚ฌ์šฉ๋  ๊ฒฝ์šฐ, Discrimiator๊ฐ€ ํ›ˆ๋ จ ๋ฐ์ดํ„ฐ๋ฅผ ๊ธฐ์–ตํ•˜๊ธฐ ๋•Œ๋ฌธ์— ์„ฑ๋Šฅ์ด ํฌ๊ฒŒ ์ €ํ•˜๋˜๋Š” ๋ฌธ์ œ๊ฐ€ ๋ฐœ์ƒํ•œ๋‹ค. ๋…ผ๋ฌธ์—์„œ๋Š” ์ด๋ฅผ ํ•ด๊ฒฐํ•˜๊ธฐ ์œ„ํ•ด ์‹ค์ œ ์ƒ˜ํ”Œ๊ณผ ๊ฐ€์งœ ์ƒ˜ํ”Œ ๋ชจ๋‘์— ๋‹ค์–‘ํ•œ ์œ ํ˜•์˜ augmentation์„ ์ ์šฉํ•˜์—ฌ G

2021๋…„ 3์›” 5์ผ
ยท
0๊ฐœ์˜ ๋Œ“๊ธ€
post-thumbnail

[DL] Object Detection

Image classification์€ DNN์— Input์œผ๋กœ ์ด๋ฏธ์ง€๋ฅผ ๋„ฃ์œผ๋ฉด, ๊ทธ ์ด๋ฏธ์ง€์— ํ•ด๋‹นํ•˜๋Š” ํด๋ž˜์Šค๋ฅผ ๋ถ„๋ฅ˜ํ•ด๋‚ด๋Š” ๋ฌธ์ œ๋ฅผ ์˜๋ฏธํ•œ๋‹ค.OD๋Š” ์ด๋ฏธ์ง€ ๋ถ„๋ฅ˜์—์„œ ์‚ฌ๋ฌผ์˜ ์œ„์น˜๋ฅผ bounding box๋กœ ์˜ˆ์ธกํ•˜๋Š” regression์ด ์ถ”๊ฐ€๋œ ํ…Œ์Šคํฌ์ด๋‹ค. ์ผ๋ฐ˜์ ์œผ๋กœ OD๋Š”

2021๋…„ 1์›” 13์ผ
ยท
0๊ฐœ์˜ ๋Œ“๊ธ€
post-thumbnail

[ML] t-SNE๋ฅผ ์ด์šฉํ•œ ์ฐจ์›์ถ•์†Œ

tSNE (t-Stochastic Neighbor Embedding)๋Š” ๊ฐ๋…๋˜์ง€ ์•Š๋Š” ๋น„์„ ํ˜• ๊ธฐ์ˆ ์ด๋ฉฐ ์ผ๋ฐ˜์ ์œผ๋กœ ๋†’์€ ์ฐจ์›์˜ ํŠน์ง•์„ ๊ฐ€์ง„ ๋ฐ์ดํ„ฐ ์ƒ˜ํ”Œ์„ ์‹œ๊ฐํ™”ํ•˜๋Š” ๋ฐ ์‚ฌ์šฉ๋œ๋‹ค. t-SNE์˜ ์ฃผ์š” ๋ชฉ์ ์€ Data Visualization์œผ๋กœ ๊ณ ์ฐจ์› ๋ฐ์ดํ„ฐ๋ฅผ 2์ฐจ์› ๋˜๋Š”

2020๋…„ 11์›” 16์ผ
ยท
0๊ฐœ์˜ ๋Œ“๊ธ€
post-thumbnail

๊นƒํ—ˆ๋ธŒ ์ž”๋””

์˜ค๋Š˜ ํƒ‘์‹ฏ๋ณด๊ณ , DB Query๋ฌธ์„ ๋งŽ์ด ์žŠ์–ด๋ฒ„๋ ธ๋‹ค ์ƒ๊ฐํ•˜์—ฌ ํ•ญ๋กœ๊ทธ์— ๊ธฐ๋กํ•˜๋Ÿฌ ๋“ค์–ด๊ฐ”๋‹ค๊ฐ€, ์ž”๋”” ์ƒ‰๊น”์ด ๋ฐ”๋€ ๊ฒƒ์„ ๋ณด์•˜๋‹ค.ํ• ๋กœ์œˆ์ด๋ผ๊ณ  ์ž”๋”” ์ƒ‰๊น”์ด ๋…ธ๋ž‘ ~ ์ฃผํ™ฉ ~. ์˜ˆ๋ป์„œ ๊ธฐ๋ก! ๐Ÿ˜œ ๐Ÿงก๐Ÿ’›ํ•™๊ธฐ ์ค‘์—” ๊ฐœ๋ฐœ์„ ์ž˜ ๋ชปํ•˜๋Š” ๊ฒƒ ๊ฐ™์•„์„œ ์•„์‰ฝ๋‹ค.์•Œ๊ณ ๋ฆฌ์ฆ˜ ๊ณต๋ถ€๋„ ๊พธ์ค€ํžˆ ํ•˜๊ณ ,

2020๋…„ 10์›” 31์ผ
ยท
3๊ฐœ์˜ ๋Œ“๊ธ€

[๋„คํŠธ์›Œํฌ] ch3

3-1. Transport-layer services ํŠธ๋žœ์ŠคํฌํŠธ ๊ณ„์ธต์€ ํ”„๋กœ์„ธ์Šค๊ฐ„ ์ •๋ณด๋ฅผ ์ „๋‹ฌ์‹œ์ผœ์ฃผ๋Š” ๊ธฐ๋Šฅ์„ ํ•œ๋‹ค. app' process๊ฐ„์˜ ๋…ผ๋ฆฌ์ ์ธ ์ •๋ณด์ „๋‹ฌ ๊ธฐ๋Šฅ์„ ์ œ๊ณตํ•œ๋‹ค. core๋ง์—์„œ๋Š” ๊ธฐ๋ณธ์ ์œผ๋กœ network layer๊นŒ์ง€ ์ฒ˜๋ฆฌ, transport laye

2020๋…„ 10์›” 25์ผ
ยท
0๊ฐœ์˜ ๋Œ“๊ธ€
post-thumbnail

[ML] Density estimation

I. Implement histogram, KDE, MLE 1. Data (weight-height.csv) The dataset given in this task consists of columns of gender, height, and weight, and ge

2020๋…„ 10์›” 12์ผ
ยท
0๊ฐœ์˜ ๋Œ“๊ธ€

Math for machine learning

๋ผ๊ทธ๋ž‘์ฃผ ์Šน์ˆ˜๋ฒ•์Šคํ„ธ๋ง ๊ทผ์‚ฌ์ •๋ณด ์ด๋ก (https://en.wikipedia.org/wiki/Information_theory์ด๋Ÿด๊ฑฐ๋ฉด ์ด๊ณผํ• ๊ฑธ

2020๋…„ 9์›” 26์ผ
ยท
0๊ฐœ์˜ ๋Œ“๊ธ€