๋ณธ๋ฌธ ๋ฐ”๋กœ๊ฐ€๊ธฐ
728x90
๋ฐ˜์‘ํ˜•

SMPL6

STAR: Sparse Trained Articulated Human Body Regressor(2020) 0. ABSTRACT 1. ํ›จ์”ฌ ๊ฐ„๊ฒฐํ•œ ๋ชจ๋ธ (๋งค๊ฐœ๋ณ€์ˆ˜์˜ ์ˆ˜ ๊ฐ์†Œ) : SMPL๋ณด๋‹ค 80% ๋” ์ž‘์€ ๋งค๊ฐœ๋ณ€์ˆ˜๋งŒ์„ ์‚ฌ์šฉ 2. ์ฒดํ˜•์— ๋”ฐ๋ผ ๋ณ€ํ•˜๋Š” ํ˜•ํƒœ ๋ฐ ํฌ์ฆˆ๋ฅผ ํ•™์Šต์‹œํ‚ด (์ฒดํ˜• ๋ฐ BMI ํ™œ์šฉ) : SMPL์˜ ๊ฒฝ์šฐ, ์ฒดํ˜•์— ๋”ฐ๋ฅธ ๋‹ค์–‘ํ•œ ํ˜•ํƒœ์˜ ๋ณ€ํ˜•์„ ๊ณ ๋ คํ•˜์ง€ ์•Š์•˜์Œ (์ฒดํ˜•์— ์ƒ๊ด€์—†์ด ๋‹ค ๋™์ผํ•œ ๊ทผ์œก ํ˜•ํƒœ๋กœ ํ‘œํ˜„) ex) ๋ˆ„๊ตฐ๊ฐ€ ํŒ”์„ ๊ตฌ๋ถ€๋ ธ์„ ๋•Œ, ๊ฐ ์ฒดํ˜•๋งˆ๋‹ค ํŒ”๊ฟˆ์น˜ ์ฃผ๋ณ€์˜ ํ”ผ๋ถ€๋‚˜ ๊ทผ์œก์˜ ํ˜•ํƒœ๊ฐ€ ๋” ๋‹ค๋ฅด๊ฒŒ ๋ณ€ํ˜•ํ•จ(๊ณจ๊ฒฉ์ด ๋” ํฐ ์‚ฌ๋žŒ, ๊ทผ์œก์ด ๋” ๋งŽ๊ฑฐ๋‚˜ ์ ์€ ์‚ฌ๋žŒ..) 3. ํ•™์Šต ๋ฐ์ดํ„ฐ๋ฅผ ๋Š˜๋ฆผ :์ผ๋ฐ˜ํ™” ์„ฑ๋Šฅ ๊ฐœ์„  ๊ฒฐ๋ก : ์†Œํ˜• ๋ชจ๋ธ์ด๋ฉฐ, ์ƒˆ๋กœ์šด ์ฒดํ˜•์— ๋Œ€ํ•ด ๋” ์ž˜ ์ผ๋ฐ˜ํ™”๋จ, SMPL ๋Œ€์ฒด ๋ชจ๋ธ๋กœ ์‚ฌ์šฉ ๊ฐ€๋Šฅ 1 INTRODUCTION ํ•ด๋‹น ๊ด€์ ˆ(๋ฌด๋ฆŽ์ด๋ฉด ๋ฌด๋ฆŽ๊นŒ์ง€๋งŒ) ์ฃผ์œ„ ๋ถ€๋ถ„๊นŒ์ง€๋งŒ ํ•™์Šต์‹œํ‚ด : ๊ธฐ์กด SMPL ๊ฐ™์€.. 2023. 8. 10.
[4์ฃผ์ฐจ] <์ฃผ์ œ ๋ณ€๊ฒฝ> ์ฒดํ˜•๋ณ„ ์˜ท์ž…ํžˆ๊ธฐ ์‚ฌ์ „ ์กฐ์‚ฌ https://github.com/lijiaman/awesome-3d-human - ๋Œ€์‹  ์ฝ”๋“œ๊ฐ€ ์—†๋Š”๊ฒŒ ๋งŽ์Œ ใ… ใ…  https://github.com/lzhbrian/Clothes-3D ๊ธฐ์กด๋Œ€๋กœ ์ฒดํ˜•์— ๋”ฐ๋ผ ์˜ท์ž…ํžˆ๊ธฐ๋ฅผ ํ•  ๊ฒƒ์ธ์ง€ (SMPL + 3D clothesํ™” + ํ•ฉ์„ฑ + ๋””ํ…Œ์ผ ) ์˜ท ์ž…์€ ์‚ฌ๋žŒ์˜ ํ˜•ํƒœ๋ฅผ ๋”ฐ์„œ ์˜ท ๋””์ž์ธ์„ ๋ฐ”๊พธ๋Š” ๊ฒƒ์œผ๋กœ ๊ฐˆ์ง€ (์ด ์นœ๊ตฌ๋Š” ์กฐ๊ธˆ ๋‹ค๋ฅธ ๋ฐฉ์•ˆ)⇒ + upgrade ๋ฒ„์ „: ์šฐ๋ฆฌ๊ฐ€ ์ง์ ‘ ์˜ท์„ ๊ทธ๋ฆฐ ๋‹ค์Œ, ์˜ท์˜ ๋””์ž์ธ๊นŒ์ง€ ๋”ฐ์„œ ์ƒˆ๋กœ ์ž…ํžˆ๋Š” ๊ฒƒ(ํฌ๋ง ์‚ฌํ•ญ) 1๋ฒˆ์œผ๋กœ ์„ ํƒํ•  ๊ฒฝ์šฐ) input์œผ๋กœ ์‚ฌ์ง„ ๋ฐ›๊ณ , ๋ชจ๋ธ ๋Œ๋ ค์„œ ์›ํ•˜๋Š” ์˜ท ์ž…ํžˆ๋Š”๊ฑธ๋กœ ๋ณด์—ฌ์ฃผ๋Š”๊ฑฐ? 2๋ฒˆ์œผ๋กœ ์„ ํƒํ•  ๊ฒฝ์šฐ) ์˜ท ๊ทธ๋ฆฌ๊ฒŒ ํ•ด์„œ ๋””์ž์ธ์„ ์ž…ํ˜€์„œ ๋ณด์—ฌ์ฃผ๋Š” ์ •๋„?๊ฐ€ ๋  ๋“ฏ (์•„์ด๋””์–ด ์ œ์‹œ์ผ ๋ฟ - ์ฐพ์ง„ ์•Š์Œ) 1. ์ฒดํ˜•๋ณ„ .. 2023. 8. 10.
Expressive Body Capture: 3D Hands, Face, and Body from a Single Image SMPL-X๋ž€? : ๋‹จ์ผ ์ด๋ฏธ์ง€๋กœ๋ถ€ํ„ฐ, ์‹ ์ฒด๋ฟ๋งŒ ์•„๋‹ˆ๋ผ, ์†๊ณผ ์–ผ๊ตด์„ ํ†ตํ•ฉ์ ์œผ๋กœ 3D ํ˜•ํƒœ์˜ ์‹ ์ฒด๋กœ ํ‘œํ˜„ํ•˜๋Š” ๋ชจ๋ธ ์ขŒ: SMPL, ์ค‘๊ฐ„: SMPL+H, ์šฐ: SMPL-X 0. ABSTRACT : 3D ์Šค์บ”์„ ์‚ฌ์šฉํ•˜์—ฌ ์ธ๊ฐ„์˜ ๋ชธ์ฒด์— ๋Œ€ํ•œ ์ƒˆ๋กœ์šด ํ†ตํ•ฉ๋œ 3D ๋ชจ๋ธ์ธ SMPL-X๋ฅผ ํ›ˆ๋ จ : SMPL์„ ํ™•์žฅํ•ด์„œ ์†๊ณผ ํ‘œ์ •๊นŒ์ง€ ๊ตฌํ˜„ํ•˜๊ณ ์ž ํ•จ : SMPL-X๋Š” ์–ผ๊ตด, ์†, ๋ชฉ, ์‹ ์ฒด ๋“ฑ ๋‹ค์–‘ํ•œ ์ธ์ฒด ํ˜•ํƒœ์™€ ์ž์„ธ๋ฅผ ํฌํ•จํ•˜๋Š” ๋งŽ์€ ๋งค๊ฐœ๋ณ€์ˆ˜๋ฅผ ๊ฐ€์ง€๊ณ  ์žˆ๋Š”๋ฐ, ์ด๋ฅผ ์ •ํ™•ํ•˜๊ฒŒ ์ถ”์ •ํ•˜๊ธฐ ์œ„ํ•ด์„œ๋Š” ๊ฐ ๋งค๊ฐœ๋ณ€์ˆ˜์— ๋Œ€ํ•œ ์ด๋ฏธ์ง€ ์ •๋ณด์™€ ๊ด€์ ˆ ์ •๋ณด๋ฅผ ๊ฒฐํ•ฉํ•ด์„œ ์ตœ์ ํ™”๋ฅผ ์ˆ˜ํ–‰ํ•ด์•ผ ํ•œ๋‹ค๋Š” ๋ฌธ์ œ์  ์กด์žฌ : ๋”ฐ๋ผ์„œ ๊ธฐ์กด SMPLify ์•Œ๊ณ ๋ฆฌ์ฆ˜(2D์—์„œ ๊ด€์ ˆ์ •๋ณด ์ถ”์ถœํ•ด์„œ ํ•™์Šต์‹œํ‚ค๋Š”)์„ ํ™œ์šฉํ•ด์„œ SMPL-X ๋ชจ๋ธ์„ ์ตœ์ ํ™”ํ•ด์„œ ์ ํ•ฉ์‹œํ‚ค๋Š” ๋ฐฉ๋ฒ•์„ ์ œ.. 2023. 8. 4.
BodyNet: Volumetric Inference of 3D Human Body Shapes BodyNet์ด๋ž€? : ๋‹จ์ผ ์ด๋ฏธ์ง€๋กœ๋ถ€ํ„ฐ 2D pose, segmentation ์ถ”์ถœ, ๋‘ ๊ฐœ์˜ ์ •๋ณด๋ฅผ ํ™œ์šฉํ•ด 3D pose๋ฅผ ํ•™์Šต, ์ดํ›„, 3๊ฐ€์ง€ ์ •๋ณด์— RGB ์ •๋ณด๊นŒ์ง€ ํ™œ์šฉํ•ด 3D์˜ ๋ถ€ํ”ผ ๊ธฐ๋ฐ˜ ์ฒดํ˜•์„ ๊ตฌ์„ฑํ•˜๋Š” Network๋ฅผ ๋งํ•จ : end to end ํ˜•์‹ 1. ์ž…๋ ฅ RGB ์ด๋ฏธ์ง€๋Š” ๋จผ์ € 2D ํฌ์ฆˆ ์ถ”์ •๊ณผ 2D ์‹ ์ฒด ๋ถ€์œ„ ์„ธ๊ทธ๋ฉ˜ํ…Œ์ด์…˜์„ ์œ„ํ•œ ํ•˜์œ„ ๋„คํŠธ์›Œํฌ๋ฅผ ํ†ต๊ณผ 2. 2D pose์™€ segmentation์„ ํ›ˆ๋ จ 3. ํ•™์Šต๋œ 2D pose์™€ Segmentation ๊ฐ€์ค‘์น˜๋ฅผ ๊ณ ์ •ํ•ด์„œ 3D pose๋ฅผ ํ›ˆ๋ จ์‹œํ‚ด 4. ์ดํ›„, ์ด์ „์˜ ๋ชจ๋“  ๋„คํŠธ์›Œํฌ ๊ฐ€์ค‘์น˜๋ฅผ ๊ณ ์ •ํ•˜๊ณ  3D ํ˜•ํƒœ network๋ฅผ ํ›ˆ๋ จ 5. ์ถ”๊ฐ€ ์žฌํ”„๋กœ์ ์…˜ ์†์‹ค๋กœ ํ˜•ํƒœ ๋„คํŠธ์›Œํฌ ํ›ˆ๋ จํ•ด์„œ ๋ถ€ํ”ผ ๊ธฐ๋ฐ˜ ํ˜•ํƒœ ์ถ”์ • ์ž‘์—…์— ๋Œ€ํ•ด ์„ธ๋ฐ€ ์กฐ์ • 6. ๊ฒฐํ•ฉ๋œ ์†.. 2023. 8. 3.
SMPLify(Keep it SMPL): Automatic Estimation of 3D Human Pose and Shape from a Single Image SMPLify[Keep it SMPL] ์ด๋ž€? : 2D CNN(Deepcut)์„ ํ™œ์šฉํ•ด ๊ด€์ ˆ ์œ„์น˜๋ฅผ ๋ฝ‘์€ ํ›„, 3D SMPL์— ์ ์šฉํ•ด 3D Mesh๋ฅผ ๊ตฌ์„ฑํ•˜๋Š” ๋ฐฉ์‹ ์š”์•ฝ(Abstract) : ์ด๋ฏธ์ง€์—์„œ ์ธ๊ฐ„์˜ 3D ํฌ์ฆˆ์™€ ํ˜•ํƒœ๋ฅผ ์ž๋™์œผ๋กœ ์ถ”์ •ํ•˜๊ณ ์ž ํ•จ : CNN ๊ธฐ๋ฒ• Deepcut์„ ํ™œ์šฉ, 3D SMPL์˜ ๊ฒฐํ•ฉ : Datasets์˜ ๊ฒฝ์šฐ, Leeds Spors, HumanEva, Human3.6M ์‚ฌ์šฉ ์ด์  ๋ฐ ํŠน์ง•(Introduction) : ์ด์ „ ๋ฐฉ์‹์˜ ๊ฒฝ์šฐ, ํฌ์ฆˆ ์ดˆ์ ์—๋งŒ ๋งž์ท„๊ณ , 3D ํ˜•ํƒœ๋ฅผ ๋ฌด์‹œํ–ˆ์Œ โžก๏ธ 2D ์ด๋ฏธ์ง€๋กœ๋ถ€ํ„ฐ ํฌ์ฆˆ์™€ ํ˜•ํƒœ๋ฅผ ๋ชจ๋‘ ํฌํ•จํ•˜๋Š” 3D ๋ฉ”์‰ฌ๋ฅผ ์ž๋™์œผ๋กœ ์ถ”์ •ํ•˜๋Š” ํ•ด๊ฒฐ์ฑ…์„ ์ œ์‹œํ•จ Deepcut ํ™œ์šฉํ•ด 2D ๊ด€์ ˆ ์ถ”์ • โ€ป DeepCut์ด๋ž€ ๐Ÿ“š [์ฐธ๊ณ ] https://arxiv.or.. 2023. 7. 31.
SMPL: A Skinned Multi-Person Linear Model ๐Ÿ’ก SMPL: A Skinned Multi-Person Linear Model ๋ชฉ์ฐจ SMPL์˜ ์ •์˜ SMPL ์—ฐ๊ตฌ ๋ชฉ์  ๊ธฐ์กด์˜ ๋ฐฉ๋ฒ• ๋ฐ ํ•œ๊ณ„ SMPL์˜ ์›๋ฆฌ์™€ ์ž‘๋™ ์ตœ์ข… DMPL SMPL(Skinned Multi-Person Linear) ์ด๋ž€? [์ฐธ๊ณ ] โžก๏ธ SMPL์— ๋Œ€ํ•œ ๊ฐ„๋‹จํ•œ ์„ค๋ช… : skinned vertex ๊ธฐ๋ฐ˜์˜ ๋ชจ๋ธ๋กœ์„œ, ๋‹ค์–‘ํ•œ ์ธ๊ฐ„์˜ ์ฒดํ˜•์„ ํ˜„์‹ค์ ์œผ๋กœ ๋‚˜ํƒ€๋‚ด๊ณ  ์ž์—ฐ์Šค๋Ÿฌ์šด ์ž์„ธ์— ๋”ฐ๋ฅธ ๋ณ€ํ˜•์„ ์ทจํ•  ์ˆ˜ ์žˆ์œผ๋ฉฐ ์—ฐ์กฐ์ง ์›€์ง์ž„์„ ๋‚˜ํƒ€๋‚ธ๋‹ค. โ€ป skinned vertex: ์Šคํ‚จ(ํ”ผ๋ถ€)์„ ์”Œ์›Œ์ง„ ๋ผˆ๊ตฌ์กฐ์— ์†ํ•œ ๋ฉ”์‹œ์˜ ์ •์  (๋ผˆ์˜ ์›€์ง์ž„์— ๋”ฐ๋ผ ๋ณ€ํ˜•๋˜๋Š” ์ •์ ) SMPL ์—ฐ๊ตฌ์˜ ๋ชฉ์  : ๋‹ค์–‘ํ•œ ์ฒดํ˜•์„ ๋Œ€ํ‘œํ•  ์ˆ˜ ์žˆ๋Š” ์‹ค์ œ์ ์ธ ์• ๋‹ˆ๋ฉ”์ด์…˜ ์ธ๊ฐ„ ์‹ ์ฒด๋ฅผ ๋งŒ๋“ค๊ณ , ์ž์—ฐ์Šค๋Ÿฌ์šด ์ž์„ธ์— ๋”ฐ๋ผ ์ž์—ฐ์Šค๋Ÿฝ๊ฒŒ ๋ณ€ํ˜•๋˜๋ฉฐ, .. 2023. 7. 28.
728x90
๋ฐ˜์‘ํ˜•