์ฝ˜ํ…์ธ ๋กœ ๊ฑด๋„ˆ๋›ฐ๊ธฐ

ImageNet10 ๋ฐ์ดํ„ฐ ์„ธํŠธ

ImageNet10 ๋ฐ์ดํ„ฐ ์„ธํŠธ๋Š” ImageNet ๋ฐ์ดํ„ฐ๋ฒ ์ด์Šค์˜ ์†Œ๊ทœ๋ชจ ํ•˜์œ„ ์ง‘ํ•ฉ์œผ๋กœ, ๋‹ค์Œ์—์„œ ๊ฐœ๋ฐœํ–ˆ์Šต๋‹ˆ๋‹ค. Ultralytics ์—์„œ ๊ฐœ๋ฐœํ–ˆ์œผ๋ฉฐ CI ํ…Œ์ŠคํŠธ, ๊ฑด์ „์„ฑ ๊ฒ€์‚ฌ ๋ฐ ํ›ˆ๋ จ ํŒŒ์ดํ”„๋ผ์ธ์˜ ๋น ๋ฅธ ํ…Œ์ŠคํŠธ๋ฅผ ์œ„ํ•ด ์„ค๊ณ„๋˜์—ˆ์Šต๋‹ˆ๋‹ค. ์ด ๋ฐ์ดํ„ฐ ์„ธํŠธ๋Š” ํ›ˆ๋ จ ์„ธํŠธ์˜ ์ฒซ ๋ฒˆ์งธ ์ด๋ฏธ์ง€์™€ ImageNet์˜ ์ฒ˜์Œ 10๊ฐœ ํด๋ž˜์Šค์˜ ์œ ํšจ์„ฑ ๊ฒ€์‚ฌ ์„ธํŠธ์˜ ์ฒซ ๋ฒˆ์งธ ์ด๋ฏธ์ง€๋กœ ๊ตฌ์„ฑ๋ฉ๋‹ˆ๋‹ค. ํ›จ์”ฌ ๋” ์ž‘์ง€๋งŒ, ์›๋ž˜ ImageNet ๋ฐ์ดํ„ฐ ์„ธํŠธ์˜ ๊ตฌ์กฐ์™€ ๋‹ค์–‘์„ฑ์„ ๊ทธ๋Œ€๋กœ ์œ ์ง€ํ•ฉ๋‹ˆ๋‹ค.

์ฃผ์š” ๊ธฐ๋Šฅ

  • ImageNet10์€ ์›๋ž˜ ๋ฐ์ดํ„ฐ ์„ธํŠธ์˜ ์ฒ˜์Œ 10๊ฐœ ํด๋ž˜์Šค๋ฅผ ๋‚˜ํƒ€๋‚ด๋Š” 20๊ฐœ์˜ ์ด๋ฏธ์ง€๊ฐ€ ํฌํ•จ๋œ ImageNet์˜ ์••์ถ• ๋ฒ„์ „์ž…๋‹ˆ๋‹ค.
  • ๋ฐ์ดํ„ฐ ์„ธํŠธ๋Š” ์ „์ฒด ์ด๋ฏธ์ง€๋„ท ๋ฐ์ดํ„ฐ ์„ธํŠธ์˜ ๊ตฌ์กฐ๋ฅผ ๋ฐ˜์˜ํ•˜๋Š” WordNet ๊ณ„์ธต ๊ตฌ์กฐ์— ๋”ฐ๋ผ ๊ตฌ์„ฑ๋ฉ๋‹ˆ๋‹ค.
  • It is ideally suited for CI tests, sanity checks, and rapid testing of training pipelines in computer vision tasks.
  • ๋ชจ๋ธ ๋ฒค์น˜๋งˆํ‚น์„ ์œ„ํ•ด ์„ค๊ณ„๋œ ๊ฒƒ์€ ์•„๋‹ˆ์ง€๋งŒ ๋ชจ๋ธ์˜ ๊ธฐ๋ณธ ๊ธฐ๋Šฅ๊ณผ ์ •ํ™•์„ฑ์„ ๋น ๋ฅด๊ฒŒ ํŒŒ์•…ํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

๋ฐ์ดํ„ฐ ์„ธํŠธ ๊ตฌ์กฐ

์ด๋ฏธ์ง€๋„ท10 ๋ฐ์ดํ„ฐ ์„ธํŠธ๋Š” ์›๋ž˜ ์ด๋ฏธ์ง€๋„ท๊ณผ ๋งˆ์ฐฌ๊ฐ€์ง€๋กœ ์›Œ๋“œ๋„ท ๊ณ„์ธต๊ตฌ์กฐ๋ฅผ ์‚ฌ์šฉํ•ด ๊ตฌ์„ฑ๋ฉ๋‹ˆ๋‹ค. ImageNet10์˜ ๊ฐ 10๊ฐœ์˜ ํด๋ž˜์Šค๋Š” ๋™์˜์–ด ์ง‘ํ•ฉ(๋™์˜์–ด ์šฉ์–ด์˜ ๋ชจ์Œ)์œผ๋กœ ์„ค๋ช…๋ฉ๋‹ˆ๋‹ค. ImageNet10์˜ ์ด๋ฏธ์ง€์—๋Š” ํ•˜๋‚˜ ์ด์ƒ์˜ ์‹ ์…‹์œผ๋กœ ์ฃผ์„์ด ์ถ”๊ฐ€๋˜์–ด ๋‹ค์–‘ํ•œ ๊ฐ์ฒด์™€ ๊ทธ ๊ด€๊ณ„๋ฅผ ์ธ์‹ํ•˜๋Š” ๋ชจ๋ธ์„ ํ…Œ์ŠคํŠธํ•˜๊ธฐ ์œ„ํ•œ ๊ฐ„๊ฒฐํ•œ ๋ฆฌ์†Œ์Šค๋ฅผ ์ œ๊ณตํ•ฉ๋‹ˆ๋‹ค.

์• ํ”Œ๋ฆฌ์ผ€์ด์…˜

ImageNet10 ๋ฐ์ดํ„ฐ ์„ธํŠธ๋Š” ์ปดํ“จํ„ฐ ๋น„์ „ ๋ชจ๋ธ๊ณผ ํŒŒ์ดํ”„๋ผ์ธ์„ ๋น ๋ฅด๊ฒŒ ํ…Œ์ŠคํŠธํ•˜๊ณ  ๋””๋ฒ„๊น…ํ•˜๋Š” ๋ฐ ์œ ์šฉํ•ฉ๋‹ˆ๋‹ค. ํฌ๊ธฐ๊ฐ€ ์ž‘์•„ ๋น ๋ฅธ ๋ฐ˜๋ณต์ด ๊ฐ€๋Šฅํ•˜๋ฏ€๋กœ ์ง€์†์ ์ธ ํ†ตํ•ฉ ํ…Œ์ŠคํŠธ์™€ ์ •์ƒ ์ƒํƒœ ํ™•์ธ์— ์ด์ƒ์ ์ž…๋‹ˆ๋‹ค. ๋˜ํ•œ ์ „์ฒด ImageNet ๋ฐ์ดํ„ฐ ์„ธํŠธ๋กœ ๋ณธ๊ฒฉ์ ์ธ ํ…Œ์ŠคํŠธ๋ฅผ ์ง„ํ–‰ํ•˜๊ธฐ ์ „์— ์ƒˆ ๋ชจ๋ธ์ด๋‚˜ ๊ธฐ์กด ๋ชจ๋ธ์˜ ๋ณ€๊ฒฝ ์‚ฌํ•ญ์„ ๋น ๋ฅด๊ฒŒ ์˜ˆ๋น„ ํ…Œ์ŠคํŠธํ•˜๋Š” ๋ฐ ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

์‚ฌ์šฉ๋ฒ•

์ด๋ฏธ์ง€ ํฌ๊ธฐ๊ฐ€ 224x224์ธ ImageNet10 ๋ฐ์ดํ„ฐ ์„ธํŠธ์—์„œ ๋”ฅ๋Ÿฌ๋‹ ๋ชจ๋ธ์„ ํ…Œ์ŠคํŠธํ•˜๋ ค๋ฉด ๋‹ค์Œ ์ฝ”๋“œ ์กฐ๊ฐ์„ ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์‚ฌ์šฉ ๊ฐ€๋Šฅํ•œ ์ธ์ˆ˜์˜ ์ „์ฒด ๋ชฉ๋ก์€ ๋ชจ๋ธ ํ•™์Šต ํŽ˜์ด์ง€๋ฅผ ์ฐธ์กฐํ•˜์„ธ์š”.

ํ…Œ์ŠคํŠธ ์˜ˆ์ œ

from ultralytics import YOLO

# Load a model
model = YOLO("yolo11n-cls.pt")  # load a pretrained model (recommended for training)

# Train the model
results = model.train(data="imagenet10", epochs=5, imgsz=224)
# Start training from a pretrained *.pt model
yolo classify train data=imagenet10 model=yolo11n-cls.pt epochs=5 imgsz=224

์ƒ˜ํ”Œ ์ด๋ฏธ์ง€ ๋ฐ ์ฃผ์„

ImageNet10 ๋ฐ์ดํ„ฐ ์„ธํŠธ์—๋Š” ์›๋ณธ ImageNet ๋ฐ์ดํ„ฐ ์„ธํŠธ์˜ ์ด๋ฏธ์ง€ ํ•˜์œ„ ์ง‘ํ•ฉ์ด ํฌํ•จ๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค. ์ด ์ด๋ฏธ์ง€๋“ค์€ ๋ฐ์ดํ„ฐ ์„ธํŠธ์˜ ์ฒ˜์Œ 10๊ฐœ ํด๋ž˜์Šค๋ฅผ ๋Œ€ํ‘œํ•˜๋„๋ก ์„ ํƒ๋˜์–ด ๋น ๋ฅธ ํ…Œ์ŠคํŠธ์™€ ํ‰๊ฐ€๋ฅผ ์œ„ํ•œ ๋‹ค์–‘ํ•˜๋ฉด์„œ๋„ ๊ฐ„๊ฒฐํ•œ ๋ฐ์ดํ„ฐ ์„ธํŠธ๋ฅผ ์ œ๊ณตํ•ฉ๋‹ˆ๋‹ค.

๋ฐ์ดํ„ฐ ์„ธํŠธ ์ƒ˜ํ”Œ ์ด๋ฏธ์ง€ ์ด ์˜ˆ์‹œ๋Š” ์ด๋ฏธ์ง€๋„ท10 ๋ฐ์ดํ„ฐ ์„ธํŠธ์— ํฌํ•จ๋œ ์ด๋ฏธ์ง€์˜ ๋‹ค์–‘์„ฑ๊ณผ ๋ณต์žก์„ฑ์„ ๋ณด์—ฌ์ฃผ๋ฉฐ, ์ปดํ“จํ„ฐ ๋น„์ „ ๋ชจ๋ธ์˜ ๊ฑด์ „์„ฑ ๊ฒ€์‚ฌ ๋ฐ ๋น ๋ฅธ ํ…Œ์ŠคํŠธ์— ์œ ์šฉํ•จ์„ ๊ฐ•์กฐํ•ฉ๋‹ˆ๋‹ค.

์ธ์šฉ ๋ฐ ๊ฐ์‚ฌ

์—ฐ๊ตฌ ๋˜๋Š” ๊ฐœ๋ฐœ ์ž‘์—…์— ImageNet10 ๋ฐ์ดํ„ฐ ์„ธํŠธ๋ฅผ ์‚ฌ์šฉํ•˜๋Š” ๊ฒฝ์šฐ, ์›๋ณธ ImageNet ๋…ผ๋ฌธ์„ ์ธ์šฉํ•ด ์ฃผ์„ธ์š”:

@article{ILSVRC15,
         author = {Olga Russakovsky and Jia Deng and Hao Su and Jonathan Krause and Sanjeev Satheesh and Sean Ma and Zhiheng Huang and Andrej Karpathy and Aditya Khosla and Michael Bernstein and Alexander C. Berg and Li Fei-Fei},
         title={ImageNet Large Scale Visual Recognition Challenge},
         year={2015},
         journal={International Journal of Computer Vision (IJCV)},
         volume={115},
         number={3},
         pages={211-252}
}

We would like to acknowledge the ImageNet team, led by Olga Russakovsky, Jia Deng, and Li Fei-Fei, for creating and maintaining the ImageNet dataset. The ImageNet10 dataset, while a compact subset, is a valuable resource for quick testing and debugging in the machine learning and computer vision research community. For more information about the ImageNet dataset and its creators, visit the ImageNet website.

์ž์ฃผ ๋ฌป๋Š” ์งˆ๋ฌธ

์ด๋ฏธ์ง€๋„ท10 ๋ฐ์ดํ„ฐ ์„ธํŠธ๋Š” ๋ฌด์—‡์ด๋ฉฐ ์ „์ฒด ์ด๋ฏธ์ง€๋„ท ๋ฐ์ดํ„ฐ ์„ธํŠธ์™€ ์–ด๋–ป๊ฒŒ ๋‹ค๋ฅธ๊ฐ€์š”?

ImageNet10 ๋ฐ์ดํ„ฐ ์„ธํŠธ๋Š” ์‹ ์†ํ•œ CI ํ…Œ์ŠคํŠธ, ๊ฑด์ „์„ฑ ๊ฒ€์‚ฌ, ํ›ˆ๋ จ ํŒŒ์ดํ”„๋ผ์ธ ํ‰๊ฐ€๋ฅผ ์œ„ํ•ด Ultralytics ์—์„œ ๋งŒ๋“  ์›๋ณธ ImageNet ๋ฐ์ดํ„ฐ๋ฒ ์ด์Šค์˜ ์••์ถ•๋œ ํ•˜์œ„ ์ง‘ํ•ฉ์ž…๋‹ˆ๋‹ค. ImageNet10์€ 20๊ฐœ์˜ ์ด๋ฏธ์ง€๋กœ๋งŒ ๊ตฌ์„ฑ๋˜์–ด ์žˆ์œผ๋ฉฐ, ImageNet์˜ ์ฒ˜์Œ 10๊ฐœ ํด๋ž˜์Šค์˜ ํ›ˆ๋ จ ๋ฐ ๊ฒ€์ฆ ์„ธํŠธ์˜ ์ฒซ ๋ฒˆ์งธ ์ด๋ฏธ์ง€๋ฅผ ๋‚˜ํƒ€๋ƒ…๋‹ˆ๋‹ค. ํฌ๊ธฐ๋Š” ์ž‘์ง€๋งŒ ์ „์ฒด ๋ฐ์ดํ„ฐ ์„ธํŠธ์˜ ๊ตฌ์กฐ์™€ ๋‹ค์–‘์„ฑ์„ ์œ ์ง€ํ•˜๋ฏ€๋กœ ๋น ๋ฅธ ํ…Œ์ŠคํŠธ์—๋Š” ์ด์ƒ์ ์ด์ง€๋งŒ ๋ชจ๋ธ ๋ฒค์น˜๋งˆํ‚น์—๋Š” ์ ํ•ฉํ•˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค.

ImageNet10 ๋ฐ์ดํ„ฐ ์„ธํŠธ๋ฅผ ์‚ฌ์šฉํ•˜์—ฌ ๋”ฅ๋Ÿฌ๋‹ ๋ชจ๋ธ์„ ํ…Œ์ŠคํŠธํ•˜๋ ค๋ฉด ์–ด๋–ป๊ฒŒ ํ•ด์•ผ ํ•˜๋‚˜์š”?

์ด๋ฏธ์ง€ ํฌ๊ธฐ๊ฐ€ 224x224์ธ ์ด๋ฏธ์ง€๋„ท10 ๋ฐ์ดํ„ฐ ์„ธํŠธ์—์„œ ๋”ฅ๋Ÿฌ๋‹ ๋ชจ๋ธ์„ ํ…Œ์ŠคํŠธํ•˜๋ ค๋ฉด ๋‹ค์Œ ์ฝ”๋“œ ์Šค๋‹ˆํŽซ์„ ์‚ฌ์šฉํ•˜์„ธ์š”.

ํ…Œ์ŠคํŠธ ์˜ˆ์ œ

from ultralytics import YOLO

# Load a model
model = YOLO("yolo11n-cls.pt")  # load a pretrained model (recommended for training)

# Train the model
results = model.train(data="imagenet10", epochs=5, imgsz=224)
# Start training from a pretrained *.pt model
yolo classify train data=imagenet10 model=yolo11n-cls.pt epochs=5 imgsz=224

์‚ฌ์šฉ ๊ฐ€๋Šฅํ•œ ์ธ์ˆ˜์˜ ์ „์ฒด ๋ชฉ๋ก์€ ๊ต์œก ํŽ˜์ด์ง€๋ฅผ ์ฐธ์กฐํ•˜์„ธ์š”.

CI ํ…Œ์ŠคํŠธ ๋ฐ ๊ฑด์ „์„ฑ ๊ฒ€์‚ฌ์— ImageNet10 ๋ฐ์ดํ„ฐ ์„ธํŠธ๋ฅผ ์‚ฌ์šฉํ•ด์•ผ ํ•˜๋Š” ์ด์œ ๋Š” ๋ฌด์—‡์ธ๊ฐ€์š”?

The ImageNet10 dataset is designed specifically for CI tests, sanity checks, and quick evaluations in deep learning pipelines. Its small size allows for rapid iteration and testing, making it perfect for continuous integration processes where speed is crucial. By maintaining the structural complexity and diversity of the original ImageNet dataset, ImageNet10 provides a reliable indication of a model's basic functionality and correctness without the overhead of processing a large dataset.

ImageNet10 ๋ฐ์ดํ„ฐ ์„ธํŠธ์˜ ์ฃผ์š” ๊ธฐ๋Šฅ์€ ๋ฌด์—‡์ธ๊ฐ€์š”?

ImageNet10 ๋ฐ์ดํ„ฐ ์„ธํŠธ์—๋Š” ๋ช‡ ๊ฐ€์ง€ ์ฃผ์š” ๊ธฐ๋Šฅ์ด ์žˆ์Šต๋‹ˆ๋‹ค:

  • ์ปดํŒฉํŠธํ•œ ํฌ๊ธฐ: 20๊ฐœ์˜ ์ด๋ฏธ์ง€๋งŒ ์žˆ์œผ๋ฉด ์‹ ์†ํ•œ ํ…Œ์ŠคํŠธ์™€ ๋””๋ฒ„๊น…์ด ๊ฐ€๋Šฅํ•ฉ๋‹ˆ๋‹ค.
  • ๊ตฌ์กฐํ™”๋œ ์กฐ์ง: ์ „์ฒด ์ด๋ฏธ์ง€๋„ท ๋ฐ์ดํ„ฐ ์„ธํŠธ์™€ ์œ ์‚ฌํ•œ ์›Œ๋“œ๋„ท ๊ณ„์ธต ๊ตฌ์กฐ๋ฅผ ๋”ฐ๋ฆ…๋‹ˆ๋‹ค.
  • CI ๋ฐ ๊ฑด์ „์„ฑ ๊ฒ€์‚ฌ: ์ง€์†์ ์ธ ํ†ตํ•ฉ ํ…Œ์ŠคํŠธ ๋ฐ ๊ฑด์ „์„ฑ ๊ฒ€์‚ฌ์— ์ด์ƒ์ ์ž…๋‹ˆ๋‹ค.
  • ๋ฒค์น˜๋งˆํ‚น์šฉ์ด ์•„๋‹™๋‹ˆ๋‹ค: ๋น ๋ฅธ ๋ชจ๋ธ ํ‰๊ฐ€์—๋Š” ์œ ์šฉํ•˜์ง€๋งŒ ๊ด‘๋ฒ”์œ„ํ•œ ๋ฒค์น˜๋งˆํ‚น์—๋Š” ์ ํ•ฉํ•˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค.

ImageNet10 ๋ฐ์ดํ„ฐ ์„ธํŠธ๋Š” ์–ด๋””์—์„œ ๋‹ค์šด๋กœ๋“œํ•  ์ˆ˜ ์žˆ๋‚˜์š”?

ImageNet10 ๋ฐ์ดํ„ฐ ์„ธํŠธ๋Š” Ultralytics GitHub ๋ฆด๋ฆฌ์Šค ํŽ˜์ด์ง€์—์„œ ๋‹ค์šด๋กœ๋“œํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ๊ตฌ์กฐ์™€ ํ™œ์šฉ์— ๋Œ€ํ•œ ์ž์„ธํ•œ ๋‚ด์šฉ์€ ImageNet10 ๋ฐ์ดํ„ฐ ์„ธํŠธ ํŽ˜์ด์ง€๋ฅผ ์ฐธ์กฐํ•˜์„ธ์š”.


๐Ÿ“… Created 11 months ago โœ๏ธ Updated 10 days ago

๋Œ“๊ธ€