Brief Review — ResT V2: Simpler, Faster and Stronger

ResTv2, Enhanced ResTv1 by Efficient Multi-Head Self-Attention v2 (EMSAv2)

Upsampling Improves Accuracy Without Increasing GFLOPs
  • ResTv2 simplifies the EMSA structure in ResTv1.
  • An upsample operation is introduced to reconstruct the lost medium- and high-frequency information caused by the downsampling operation.

Outline

  1. Efficient Multi-Head Self-Attention v2 (EMSAv2)
  2. Results

1. Efficient Multi-Head Self-Attention v2 (EMSAv2)

Left: Efficient Multi-Head Self-Attention (EMSA), Right: EMSAv2

1.1. EMSA in ResTv1

  • To compress memory, x is reshaped to its 2D size and then are downsampled by a depth-wise convolution to reduce the height and width:

1.2. EMSAv2

  • An upsampling operation is introduced on the values directly. There are many upsampling strategies, such as “nearest”, “bilinear”, “pixel-shuffle”, etc.
  • All of them can improve the model’s performance, but “pixel-shuffle” works better.
  • Surprisingly, this “downsample-upsample” combination in EMSAv2 happens to build an independent convolution branch, which can efficiently reconstruct the lost information with fewer extra parameters and computation costs.
  • And the multi-head interaction module of the self-attention branch in EMSAv2 will decrease the actual inference speed of EMSAv2, although it can increase the final performance.
  • Therefore, it is removed for faster speed under default settings. However, if the head dimension is small (e.g., dk=64 or smaller), the multi-head interaction module will make a difference.

1.3. ResTv2 Model Variants

  • Different ResTv2 (T/S/B/L) variants are built based on EMSAv2.
  • ResTv2-T/B/L, to be of similar complexities to Swin-T/S/B:
ResTv2 Model Variants

2. Results

2.1. ImageNet

Classification accuracy on ImageNet-1k.

2.2. Ablation Study

Ablation Study
  • (a) Upsampling Targets: Upsampling V works best.
  • (b) Upsampling Strategies: Pixel-shuffle operation obtains much stronger feature extraction capabilities with a few parameters and FLOPs increase.
  • (c) ConvNet or EMSA?: The “downsampling-upsampling” pipeline in EMSAv2 can constitute a complete ConvNet block for extracting features.
  • (d) Positional Embedding: PE can still improve the performance, but not that obvious as ResTv1.
Different MSA Variants
Multi-Head Interaction Module (MHIM)

2.3. Downstream

Object detection results of fine-tuning styles on COCO val2017 with ResTv2-T using Mask R-CNN.
  • Win: The recent popular one is window-style, which constrained part or all MSA modules of ViTs into a fixed window to save computation overhead. However, performing all MSA into a limited-sized window will lose the MSA’s long-range dependency ability.
  • CWin: To alleviate this issue, we add a 7×7 depth-wise convolution layer after the last block in each stage to enable information to communicate across windows.
  • HWin: In addition, MViTv2 [20] provides a hybrid approach to integrate window information, i.e., computes MSA within a window in all but the last blocks in each stage that feed into FPN.
  • Window sizes in Win, CWin, and HWin are set as [64, 32, 16, 8] for the four stages.
COCO object detection and segmentation results using Mask-RCNN.
ADE20K validation results using UPerNet.

Reference

1.1. Image Classification

My Other Previous Paper Readings

--

--

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store