--- id: P-REINFORCE-AI-CV category: "[[10_Wiki/๐Ÿ’ก Topics/AI]]" confidence_score: 0.99 tags: [Computer Vision, Deep Learning, Image Processing, Object Detection] last_reinforced: 2026-04-20 --- # [[Computer-Vision]] (์ปดํ“จํ„ฐ ๋น„์ „) ## ๐Ÿ“Œ ํ•œ ์ค„ ํ†ต์ฐฐ (The Karpathy Summary) > "๊ธฐ๊ณ„์—๊ฒŒ ๋ˆˆ์„ ๋ถ€์—ฌํ•˜๋Š” ๊ธฐ์ˆ ." ํ”ฝ์…€ ๋ฐ์ดํ„ฐ์—์„œ ํŒจํ„ด์„ ์ฐพ์•„๋‚ด๊ณ , ๊ทธ๊ฒƒ์ด '๊ณ ์–‘์ด'์ธ์ง€ '๋ณดํ–‰์ž'์ธ์ง€, ์•„๋‹ˆ๋ฉด '์•”์„ธํฌ'์ธ์ง€ ํ•ด์„ํ•˜๋Š” AI์˜ ์‹œ๊ฐ ์‹œ์Šคํ…œ์ด๋‹ค. ## ๐Ÿ“– ๊ตฌ์กฐํ™”๋œ ์ง€์‹ (Synthesized Content) - **Image Recognition & Classification**: - ์‚ฌ์ง„์„ ๋ณด๊ณ  ๋ฌด์—‡์ธ์ง€ ๋ ˆ์ด๋ธ”๋งํ•˜๋Š” ๊ธฐ์ดˆ ๋‹จ๊ณ„. CNN(Convolutional Neural Networks)์˜ ๋“ฑ์žฅ์œผ๋กœ ํ˜๋ช…์ ์ธ ๋ณ€ํ™”๊ฐ€ ์ผ์–ด๋‚ฌ๋‹ค. - **Object Detection & Segmentation**: - ํ™”๋ฉด ์•ˆ์˜ ์‚ฌ๋ฌผ ์œ„์น˜๋ฅผ ๋ฐ•์Šค๋กœ ํ‘œ์‹œ(Detection)ํ•˜๊ฑฐ๋‚˜, ํ”ฝ์…€ ๋‹จ์œ„๋กœ ๊ฒฝ๊ณ„๋ฅผ ์น ํ•˜๋Š”(Segmentation) ์ •๋ฐ€ ์ž‘์—…. ์ž์œจ์ฃผํ–‰์˜ ํ•ต์‹ฌ์ด๋‹ค. - **Vision Transformers (ViT)**: - ์ตœ๊ทผ NLP์—์„œ ์“ฐ์ด๋Š” 'Attention' ๊ธฐ๋ฒ•์„ ์ด๋ฏธ์ง€ ์ฒ˜๋ฆฌ์— ๋„์ž…ํ•˜์—ฌ, ๊ธฐ์กด CNN์˜ ํ•œ๊ณ„๋ฅผ ๋„˜์–ด์„œ๋Š” ์„ฑ๋Šฅ์„ ๋ณด์—ฌ์ฃผ๊ณ  ์žˆ๋‹ค. ## โš ๏ธ ๋ชจ์ˆœ ๋ฐ ์—…๋ฐ์ดํŠธ (RL Update) - ์ปดํ“จํ„ฐ ๋น„์ „์€ ๋น›์˜ ์กฐ๊ฑด, ๊ฐ€๋ฆผ(Occlusion), ์•ต๊ธ€์˜ ๋ณ€ํ™”์— ์—ฌ์ „ํžˆ ์ทจ์•ฝํ•œ ๋ฉด์ด ์žˆ๋‹ค. ์ด๋ฅผ ๊ทน๋ณตํ•˜๊ธฐ ์œ„ํ•ด ๋‹ค๊ฐ๋„ ์นด๋ฉ”๋ผ์™€ ๋ผ์ด๋‹ค(LiDAR) ๋ฐ์ดํ„ฐ๋ฅผ ํ•ฉ์น˜๋Š” '์„ผ์„œ ํ“จ์ „(Sensor Fusion)' ๊ธฐ์ˆ ์ด ํ™œ๋ฐœํžˆ ์—ฐ๊ตฌ๋˜๊ณ  ์žˆ๋‹ค. ## ๐Ÿ”— ์ง€์‹ ์—ฐ๊ฒฐ (Graph) - Related: [[Autonomous-Vehicle-Path-Planning]] , [[Robotic Manipulation]] - Foundation: [[Information Theory]]