The implementation of Isotropic architecture

[buptxiaofeng]

Hi, thanks for sharing this impressive work. The paper mentioned two architectures, Isotropic one and pyramid one. I noticed that in the code, this is a reduce_ratios, and this reduce_ratios are used by a avg_pooling operation to calculate before building the graph. I am wondering whether all I need to do is setting this reduce_ratios to [1,1,1,1] if I want to implement the Isotropic architecture. Thanks.

self.n_blocks = sum(blocks)
channels = opt.channels
reduce_ratios = [4, 2, 1, 1]
dpr = [x.item() for x in torch.linspace(0, drop_path, self.n_blocks)]
num_knn = [int(x.item()) for x in torch.linspace(k, k, self.n_blocks)]

[iamhankai]

Yes, set this reduce_ratios to [1,1,1,1].

[HaniItani]

Hello @iamhankai,

Shouldn't the downsampling be also disabled for the isotropic architecture so that the number of tokens is constant throughout the model?

[iamhankai]

Yes, it doesn't need downsampling in the isotropic architecture.

[0xf21]

for i in range(len(blocks)):

        if i > 0:

            self.backbone.append(Downsample(channels[i-1], channels[i]))

To implement the Isotropic architecture, I need to do is setting reduce_ratios to [1,1,1,1] and removing this Downsample, right?

[iamhankai]

Yes. The Isotropic code will be released soon.

[iamhankai]

Yes. The Isotropic code will be released soon.

已发布isotropic ViG的代码：https://github.com/huawei-noah/Efficient-AI-Backbones/tree/master/vig_pytorch