๋ณธ๋ฌธ ๋ฐ”๋กœ๊ฐ€๊ธฐ
728x90
๋ฐ˜์‘ํ˜•

Deep Learning78

[Standford_cs231n] Lecture 1 ) Introduction to Convolutional Neural Networks for Visual Recognition ์ปดํ“จํ„ฐ ๋น„์ „์˜ ์—ญ์‚ฌ 1. ์ปดํ“จํ„ฐ ๋น„์ „์ด๋ž€ ⇒ ์ตœ๊ทผ ์ธํ„ฐ๋„ท ํŠธ๋ž˜ํ”ฝ ์ค‘ 80%๊ฐ€ ๋น„๋””์˜ค ๋ฐ์ดํ„ฐ์ผ ๋งŒํผ ์—„์ฒญ๋‚œ ์–‘์˜ ์‹œ๊ฐ์  ๋ฐ์ดํ„ฐ๋“ค์ด ์Ÿ์•„์ ธ ๋‚˜์˜ค๋Š” ๊ฒƒ์„ ์œ ํŠœ๋ธŒ๋งŒ ๋ด๋„ ๋Š๋‚„ ์ˆ˜ ์žˆ์Œ. ⇒ ์ด ๋ฐ์ดํ„ฐ๋“ค์„ ํšจ๊ณผ์ ์œผ๋กœ ์ดํ•ดํ•˜๊ณ  ๋ถ„์„ํ•ด์„œ, ์ปดํ“จํ„ฐ๋กœ ํ•˜์—ฌ๊ธˆ ์ธ๊ฐ„์˜ ์‹œ๊ฐ์ ์ธ ์ธ์‹ ๋Šฅ๋ ฅ์„ ๊ฐ€์งˆ ์ˆ˜ ์žˆ๋„๋ก ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ๊ฐœ๋ฐœํ•˜๊ธฐ ์‹œ์ž‘ํ–ˆ๊ณ , ์ด๊ฒƒ์„ ์ปดํ“จํ„ฐ ๋น„์ „์ด๋ผ ํ•จ. 2. ์ปดํ“จํ„ฐ ๋น„์ „์˜ ์—ญ์‚ฌ 1950s Hubel๊ณผ Wiesel ์ƒ๋ฌผ์˜ ์‹œ๊ฐ์  ๋งค์ปค๋‹ˆ์ฆ˜์„ ์ฐพ๊ณ ์ž ๊ณ ์–‘์ด ๋‡Œ์— ์ „๊ทน์„ ๊ฝ‚์•„ ์‹คํ—˜ ์ง„ํ–‰ ๊ณ ์–‘์ด์—๊ฒŒ ์–ด๋– ํ•œ ์‹œ๊ฐ์  ์ž๊ทน์„ ์ฃผ์–ด์•ผ ๊ณ ์–‘์ด์˜ ๋‡Œ์˜ 1์ฐจ ์‹œ๊ฐ ํ”ผ์งˆ์˜ ๋‰ด๋Ÿฐ๋“ค์ด ๊ฒฉ๋ ฌํ•˜๊ฒŒ ๋ฐ˜์‘ํ• ์ง€์— ๋Œ€ํ•ด ์‹คํ—˜ ์ง„ํ–‰ ์‹œ๊ฐ์  input์˜ edges๊ฐ€ ์›€์ง์ผ ๋•Œ ๋ฐ˜์‘ํ•˜๋Š” ๋‹จ์ˆœํ•œ ์„ธํฌ์— ์ดˆ์ ์„ ๋‘  ⇒ "์‹œ๊ฐ ์ฒ˜๋ฆฌ๋Š” edges์™€ ๊ฐ™์€ ๋‹จ์ˆœํ•œ .. 2023. 7. 7.
EfficientNet 1. Intro ์ด์ „๊นŒ์ง€๋Š” depth, width, size ์ค‘ ํ•˜๋‚˜๋งŒ scale ํ•˜๋Š” ๊ฒƒ์„ ์ฃผ๋กœ ๋‹ค๋ฃธ⇒ ๋” ๋‚˜์€ ์ •ํ™•๋„ ํ˜น์€ ํšจ์œจ์„ฑ์œผ๋กœ convnet์„ scale up ํ•˜๋Š” ๋ฐฉ๋ฒ•์ด ์—†์„๊นŒ?์—์„œ ๋‚˜์˜จ ์นœ๊ตฌ depth, width, size ์„ธ ๊ฐ€์ง€ ๊ท ํ˜•์„ ์ž˜ ๋งž์ถ”๋Š” ๊ฒƒ์ด ์ค‘์š”ํ•จ ์ƒ์ˆ˜ ๋น„์œจ๋กœ ์„ธ ๊ฐ€์ง€๋ฅผ ๊ฐ๊ฐ scalingํ•˜๋ฉด ๋œ๋‹ค๋Š” ๊ฒƒ์„ ์•Œ๊ฒŒ ๋จ 2. Model Scaling Convnet์˜ ์ •ํ™•๋„๋ฅผ ๋†’์ผ ๋•Œ ์ž˜ ์งœ์—ฌ์ง„ ๋ชจ๋ธ์„ ์ฐพ๋Š” ๋ฐฉ๋ฒ•๋„ ์žˆ์ง€๋งŒ, ๊ธฐ์กด ๋ชจ๋ธ์„ ๋ฐ”ํƒ•์œผ๋กœ ๋ณต์žก๋„๋ฅผ ๋†’์ด๋Š” ๋ฐฉ๋ฒ•๋„ ๋งŽ์ด ์‚ฌ์šฉ depth scaling: layer์˜ ๊ฐœ์ˆ˜๋ฅผ ๋†’์—ฌ์คŒ ex) ResNet width scaling: channel(ํ•„ํ„ฐ) ๊ฐœ์ˆ˜๋ฅผ ๋†’์—ฌ์คŒ ex) MobileNet, ShuffleNet resolution sc.. 2023. 7. 7.
cGAN/Pix2Pix 1. GAN 2. cGAN ์–ด๋–ค ์ˆซ์ž๋ฅผ ๋งŒ๋“ค์–ด๋‚ผ์ง€์— ๋Œ€ํ•œ ์ •๋ณด๋ฅผ ๋„ฃ์–ด์ฃผ๋Š” ๊ฒƒ(์–ด๋–ค ํด๋ž˜์Šค์— ํ•ด๋‹นํ•˜๋Š”์ง€์— ๋Œ€ํ•œ ์ •๋ณด) ex) 7์„ ๋งŒ๋“ค๊ณ ์ž ํ•œ๋‹ค๋ฉด condition vector์— 7์„ ๋„ฃ์–ด์ฃผ๊ณ , z(noise)์—๋Š” ๋žœ๋คํ•˜๊ฒŒ ์ƒ˜ํ”Œ๋งํ•ด์„œ 7์ด๋ผ๋Š” ์˜๋ฏธ๋ฅผ ๊ฐ€์ง€๋Š” ๋žœ๋คํ•œ ํ˜•ํƒœ๋ฅผ ๋งŒ๋“ค์–ด์คŒ 3. Pix2Pix image to image translation : ์ด๋ฏธ์ง€์˜ ํŠน์ • ์–‘์ƒ์„ ๋‹ค๋ฅธ ์–‘์ƒ์œผ๋กœ ๋ฐ”๊ฟ”์ฃผ๋Š” ๊ฒƒ์„ ์˜๋ฏธ ex) ์†๊ทธ๋ฆผ → ์‹ค์ œ ์‚ฌ์ง„์œผ๋กœ translation ์ด๋ฏธ์ง€ ์ž์ฒด๋ฅผ condition์œผ๋กœ ๋ฐ›์•„๋ฒ„๋ฆผ (์ด๋ฏธ์ง€ ์ž์ฒด๊ฐ€ ์ •๋ณด๊ฐ€ ๋˜๋Š” ๊ฒƒ์ž„) ์ฆ‰, ์ด๋ฏธ์ง€๊ฐ€ ๋“ค์–ด์™”์„ ๋•Œ, ๊ฑฐ๊ธฐ์— ๋ถ€ํ•ฉํ•˜๋Š” output์˜ ํ˜•ํƒœ๋กœ ๋งŒ๋“ค์–ด์คŒ noise vector z๋ฅผ ์•ˆ์”€ pixel ์„ ์ž…๋ ฅ์œผ๋กœ ๋ฐ›์•„ pixel์„ ์˜ˆ์ธกํ•จ (์ด๋ฏธ์ง€๋ฅผ ์ด๋ฏธ.. 2023. 7. 7.
R-CNN 1. Intro R-CNN 'Rich feature hierarchies for accurate object detection and semantic segmentation'. R-CNN์€ region proposals์™€ CNN์ด ๊ฒฐํ•ฉ๋œ Regions with CNN์˜ ์•ฝ์ž๋กœ ์ง€์นญ (1) region proposals๋กœ object ์œ„์น˜๋ฅผ ์•Œ์•„๋‚ด๊ณ , ์ด๋ฅผ CNN์— ์ž…๋ ฅํ•˜์—ฌ class๋ฅผ ๋ถ„๋ฅ˜. (2) Larger data set์œผ๋กœ ํ•™์Šต๋œ pre-trained CNN์„ fine-tunning. 2. Overall architecture ์ž…๋ ฅ ์ด๋ฏธ์ง€์— Selective Search ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ์ ์šฉํ•˜์—ฌ bounding box(region proposal) 2000๊ฐœ๋ฅผ ์ถ”์ถœ. ์ถ”์ถœ๋œ bounding box๋ฅผ w.. 2023. 7. 6.
GAN: Generative Adversarial Nets 0. Abstract ๋ณธ๋ฌธ We propose a new framework for estimating generative models via an adversarial process, in which we simultaneously train two models: a generative model G that captures the data distribution, and a discriminative model D that estimates the probability that a sample came from the training data rather than G. The training procedure for G is to maximize the probability of D making a m.. 2023. 7. 6.
AE AutoEncoder ์ž…๋ ฅ์ด ๋“ค์–ด์™”์„ ๋•Œ, ํ•ด๋‹น ์ž…๋ ฅ ๋ฐ์ดํ„ฐ๋ฅผ ์ตœ๋Œ€ํ•œ ์••์ถ• ์‹œํ‚จ ํ›„, ์••์ถ• ์‹œํ‚จ ๋ฐ์ดํ„ฐ๋ฅผ ๋ณธ๋ž˜์˜ ์ž…๋ ฅ ํ˜•ํƒœ๋กœ ๋ณต์›์‹œํ‚ค๋Š” ์‹ ๊ฒฝ๋ง ์••์ถ•ํ•˜๋Š” ๋ถ€๋ถ„์„ encoder ๋ณต์›ํ•˜๋Š” ๋ถ€๋ถ„์„ decoder ์••์ถ•๊ณผ์ •์—์„œ ์ถ”์ถœํ•œ ์˜๋ฏธ์žˆ๋Š” ๋ฐ์ดํ„ฐ๋ฅผ latent vector AutoEncoder์˜ ์ˆ˜์‹๊ณผ ํ•™์Šต ๋ฐฉ๋ฒ• ์ˆ˜์‹ Input Data๋ฅผ Encoder Network์— ํ†ต๊ณผ์‹œ์ผœ ์••์ถ•๋œ z๊ฐ’์„ ์–ป์Œ ์••์ถ•๋œ z vector๋กœ๋ถ€ํ„ฐ Input Data์™€ ๊ฐ™์€ ํฌ๊ธฐ์˜ ์ถœ๋ ฅ ๊ฐ’์„ ์ƒ์„ฑ ์ด๋•Œ Loss๊ฐ’์€ ์ž…๋ ฅ๊ฐ’ x์™€ Decoder๋ฅผ ํ†ต๊ณผํ•œ y๊ฐ’์˜ ์ฐจ์ด ํ•™์Šต ๋ฐฉ๋ฒ• Decoder Network๋ฅผ ํ†ต๊ณผํ•œ Output layer์˜ ์ถœ๋ ฅ ๊ฐ’์€ Input๊ฐ’์˜ ํฌ๊ธฐ์™€ ๊ฐ™์•„์•ผ ํ•จ(๊ฐ™์€ ์ด๋ฏธ์ง€๋ฅผ ๋ณต์›ํ•œ๋‹ค๊ณ  ์ƒ๊ฐํ•˜๋ฉด ๋จ) ์ด๋•Œ ํ•™์Šต์„ ์œ„ํ•ด์„œ.. 2023. 7. 6.
728x90
๋ฐ˜์‘ํ˜•