DeiT-B.

Metaopen-source86M paramsVision TransformerApache 2.0

83.1% on ImageNet without external data.

§ 01 · Benchmarks

Every benchmark DeiT-B has a recorded score for.

#	Benchmark	Area · Task	Metric	Value	Rank	Date	Source
01	publaynet-val	Computer Vision · Document Layout Analysis	title	0.9%	#9/14	2020-12-23	source ↗
02	publaynet-val	Computer Vision · Document Layout Analysis	text	0.9%	#9/14	2020-12-23	source ↗
03	publaynet-val	Computer Vision · Document Layout Analysis	figure	1.0%	#10/14	2020-12-23	source ↗
04	publaynet-val	Computer Vision · Document Layout Analysis	overall	0.9%	#10/14	2020-12-23	source ↗
05	publaynet-val	Computer Vision · Document Layout Analysis	list	0.9%	#11/14	2020-12-23	source ↗
06	publaynet-val	Computer Vision · Document Layout Analysis	table	1.0%	#12/15	2020-12-23	source ↗
07	ImageNet-1K	Computer Vision · Image Classification	top-1-accuracy	83.1%	#13/20	—	source ↗
08	rvl-cdip	Computer Vision · Document Image Classification	accuracy	90.3%	#34/35	2020-12-23	source ↗

Rank column shows this model’s position vs all other models scored on the same benchmark + metric (competitors after the slash). #1 in red means current SOTA. Sorted by rank, then newest result.

§ 02 · Strengths by area

Where DeiT-B actually performs.

Computer Vision

benchmarks

avg rank #13.5

§ 03 · Papers

1 paper with results for DeiT-B.

2020-12-23· Computer Vision· 7 results
Training data-efficient image transformers & distillation through attention

§ 04 · Related models

Other Meta models scored on Codesota.

DeiT-B Distilled

86M params · 2 results · 1 SOTA

400B total / 17B active (128 experts) params · 6 results

Llama 3.1 70B

4 results

Code Llama 34B

Unknown params · 2 results

ConvNeXt V2 Huge

650M params · 2 results

CodeLlama 70B

70B params · 1 result

§ 05 · Sources & freshness

Where these numbers come from.

papers-with-code

results

meta-research

result

7 of 8 rows marked verified.