Neural Architecture Transfer

Open paper
Benchmark Model Rank Results
architecture-search-on-cifar-10-imageNAT-M4#1Percentage error: 1.6Params: 6.9MFLOPS: 468M
architecture-search-on-cifar-10-imageNAT-M3#2Percentage error: 1.8Params: 6.2MFLOPS: 392M
architecture-search-on-cifar-10-imageNAT-M2#7Percentage error: 2.1Params: 4.6MFLOPS: 291M
architecture-search-on-cifar-10-imageNAT-M1#15Percentage error: 2.6Params: 4.3MFLOPS: 232M
fine-grained-image-classification-on-fgvcNAT-M4#33Accuracy: 90.8%FLOPS: 581MPARAMS: 5.3M
fine-grained-image-classification-on-fgvcNAT-M3#34Accuracy: 90.1%FLOPS: 388MPARAMS: 5.1M
fine-grained-image-classification-on-fgvcNAT-M2#36Accuracy: 89.0%FLOPS: 235MPARAMS: 3.4M
fine-grained-image-classification-on-fgvcNAT-M1#38Accuracy: 87.0%FLOPS: 175MPARAMS: 3.2M
fine-grained-image-classification-on-food-101NAT-M4#9Accuracy: 89.4FLOPS: 361MPARAMS: 4.5M
fine-grained-image-classification-on-food-101NAT-M3#10Accuracy: 89.0FLOPS: 299MPARAMS: 3.9M
fine-grained-image-classification-on-food-101NAT-M2#11Accuracy: 88.5FLOPS: 266MPARAMS: 4.1M
fine-grained-image-classification-on-food-101NAT-M1#12Accuracy: 87.4FLOPS: 198MPARAMS: 3.1M
fine-grained-image-classification-on-oxfordNAT-M4#11Accuracy: 98.3%FLOPS: 400MPARAMS: 4.2M
fine-grained-image-classification-on-oxfordNAT-M3#13Accuracy: 98.1%FLOPS: 250MPARAMS: 3.7M
fine-grained-image-classification-on-oxfordNAT-M2#14Accuracy: 97.9%FLOPS: 195MPARAMS: 3.4M
fine-grained-image-classification-on-oxfordNAT-M1#24FLOPS: 152MPARAMS: 3.3M
fine-grained-image-classification-on-oxford-1NAT-M1#13FLOPS: 160MPARAMS: 4.0M
fine-grained-image-classification-on-oxford-2NAT-M4#7Accuracy: 94.3Top-1 Error Rate: 5.7%FLOPS: 744MPARAMS: 8.5M
fine-grained-image-classification-on-oxford-2NAT-M3#8Accuracy: 94.1Top-1 Error Rate: 5.9%FLOPS: 471MPARAMS: 5.7M
fine-grained-image-classification-on-oxford-2NAT-M2#9Accuracy: 93.5Top-1 Error Rate: 6.5%FLOPS: 306MPARAMS: 5.5M
fine-grained-image-classification-on-stanfordNAT-M4#52Accuracy: 92.9%FLOPS: 369MPARAMS: 3.7M
fine-grained-image-classification-on-stanfordNAT-M3#56Accuracy: 92.6%FLOPS: 289MPARAMS: 3.5M
fine-grained-image-classification-on-stanfordNAT-M2#59Accuracy: 92.2%FLOPS: 222MPARAMS: 2.7M
fine-grained-image-classification-on-stanfordNAT-M1#61Accuracy: 90.9%FLOPS: 165MPARAMS: 2.4M
image-classification-on-cifar-10NAT-M4#41Percentage correct: 98.4Top-1 Accuracy: 98.4Parameters: 6.9M
image-classification-on-cifar-10NAT-M3#49Percentage correct: 98.2Top-1 Accuracy: 98.2Parameters: 6.2M
image-classification-on-cifar-10NAT-M2#63Percentage correct: 97.9Top-1 Accuracy: 97.9Parameters: 4.6M
image-classification-on-cifar-10NAT-M1#85Percentage correct: 97.4Top-1 Accuracy: 97.4Parameters: 4.3M
image-classification-on-cifar-100NAT-M4#39Percentage correct: 88.3PARAMS: 9.0M
image-classification-on-cifar-100NAT-M3#41Percentage correct: 87.7PARAMS: 7.8M
image-classification-on-cifar-100NAT-M2#43Percentage correct: 87.5PARAMS: 6.4M
image-classification-on-cifar-100NAT-M1#57Percentage correct: 86.0PARAMS: 3.8M
image-classification-on-cinic-10NAT-M3#2Accuracy: 94.3FLOPS: 501MPARAMS: 8.1M
image-classification-on-cinic-10NAT-M2#3Accuracy: 94.1FLOPS: 411MPARAMS: 6.2M
image-classification-on-cinic-10NAT-M1#4Accuracy: 93.4FLOPS: 317MPARAMS: 4.6M
image-classification-on-flowers-102NAT-M4#23Accuracy: 98.3%FLOPS: 400MPARAMS: 4.2M
image-classification-on-flowers-102NAT-M3#27Accuracy: 98.1%FLOPS: 250MPARAMS: 3.7M
image-classification-on-flowers-102NAT-M2#28Accuracy: 97.9%FLOPS: 195MPARAMS: 3.4M
image-classification-on-flowers-102NAT-M1#50FLOPS: 152MPARAMS: 3.3M
image-classification-on-imagenetNAT-M4#666Top 1 Accuracy: 80.5%Number of params: 9.1M
image-classification-on-stl-10NAT-M4#3Percentage correct: 97.9FLOPS: 573MPARAMS: 7.5M
image-classification-on-stl-10NAT-M3#4Percentage correct: 97.8FLOPS: 436MPARAMS: 7.5M
image-classification-on-stl-10NAT-M2#6Percentage correct: 97.2FLOPS: 303MPARAMS: 5.1M
image-classification-on-stl-10NAT-M1#8Percentage correct: 96.7FLOPS: 240MPARAMS: 4.4M
neural-architecture-search-on-cifar-10NAT-M4#1Top-1 Error Rate: 1.6%Parameters: 6.9MFLOPS: 468M…
neural-architecture-search-on-cifar-10NAT-M3#4Top-1 Error Rate: 1.8%Parameters: 6.2MFLOPS: 392M…
neural-architecture-search-on-cifar-10NAT-M2#7Top-1 Error Rate: 2.1%Parameters: 4.6MFLOPS: 291M…
neural-architecture-search-on-cifar-10NAT-M1#24Top-1 Error Rate: 2.6%Parameters: 4.3MFLOPS: 232M…
neural-architecture-search-on-cifar-100-1NAT-M4#2Percentage Error: 11.7FLOPS: 796MPARAMS: 9.0M
neural-architecture-search-on-cifar-100-1NAT-M3#3Percentage Error: 12.3FLOPS: 492MPARAMS: 7.8M
neural-architecture-search-on-cifar-100-1NAT-M2#4Percentage Error: 12.5FLOPS: 398MPARAMS: 6.4M
neural-architecture-search-on-cifar-100-1NAT-M1#6Percentage Error: 14.0FLOPS: 261MPARAMS: 3.8M
neural-architecture-search-on-food-101NAT-M4#1Accuracy (%): 89.4FLOPS: 361MPARAMS: 4.5M
neural-architecture-search-on-food-101NAT-M3#2Accuracy (%): 89.0FLOPS: 299MPARAMS: 3.9M
neural-architecture-search-on-food-101NAT-M2#3Accuracy (%): 88.5FLOPS: 266MPARAMS: 4.1M
neural-architecture-search-on-food-101NAT-M1#4Accuracy (%): 87.4FLOPS: 198MPARAMS: 3.1M
neural-architecture-search-on-imagenetNAT-M4#13Top-1 Error Rate: 19.5Accuracy: 80.5Params: 9.1MMACs: 600M
neural-architecture-search-on-imagenetNAT-M3#22Top-1 Error Rate: 20.1Accuracy: 79.9Params: 9.1MMACs: 490M
neural-architecture-search-on-imagenetNAT-M2#41Top-1 Error Rate: 21.4Accuracy: 78.6Params: 7.7MMACs: 312M
neural-architecture-search-on-imagenetNAT-M1#59Top-1 Error Rate: 22.5Accuracy: 77.5Params: 6.0MMACs: 225M

Rank counts only results with a code link.