MODEL_ZOO.md

AMD Model Zoo

Something-Something V2

Method	Extra Data	Backbone	Epoch	#Frame	Pre-train	Fine-tune	Top-1	Top-5
AMD	no	ViT-S	800	16x2x3	log/checkpoint	log/checkpoint	70.2	92.5
AMD	no	ViT-B	800	16x2x3	log/checkpoint	log/checkpoint	73.3	94.0

Kinetics-400

Method	Extra Data	Backbone	Epoch	#Frame	Pre-train	Fine-tune	Top-1	Top-5
AMD	no	ViT-S	800	16x5x3	log/checkpoint	log/checkpoint	80.1	94.5
AMD	no	ViT-B	800	16x5x3	log/checkpoint	log/checkpoint	82.2	95.3

ImageNet-1K

Method	Extra Data	Backbone	Epoch	Pre-train	Fine-tune	Top-1
AMD	no	ViT-S	800	log/checkpoint	log/checkpoint	82.1
AMD	no	ViT-B	800	log/checkpoint	log/checkpoint	84.6

Note:

• We report the results of AMD finetuned with I3D dense sampling on Kinetics-400 and TSN uniform sampling on Something-Something V2, respectively.
• #Frame = #input_frame x #clip x #crop.
• #input_frame means how many frames are input for model during the test phase.
• #crop means spatial crops (e.g., 3 for left/right/center crop).
• #clip means temporal clips (e.g., 5 means repeted temporal sampling five clips with different start indices).