论文阅读笔记：From Image-level to Pixel-level Labeling with Convolutional Networks

Introduce

Task

Weakly supervised learning for image semantic segmentation, only use image class label.

Contribution

• Combine MIL(Multiple Instance Learning) with classification CNN
• experiments is state of the art

Framework

Train

For input image $I:3\times h \times w$I:3×h×w, pass a backbone(i.e. Overfeat + Segmentation Net), output feature maps $Y:(|C|+1) \times h^{o} \times w^{o}$Y:(∣C∣+1)×ho×wo, then $Y$Y pass a LSE(Log-Sum-Exp) pooling, output $s:(|C|+1) \times 1 \times 1$s:(∣C∣+1)×1×1. Finally compute a softmax cross entrophy loss for $s$s, backpropagation gradients to train backbone.

Inference

$p_{i,j}(k)$pi,j​(k) be the $Y$Y for location $(i,j)$(i,j) and $k^{th}$kth class label. ILP $p(k)$p(k) be the $s$s by softmax.
$\widehat{y}_{i,j}=p_{i,j}(k|I) \times p(k|I)$y​i,j​=pi,j​(k∣I)×p(k∣I)
Finally, $\widehat{y}_{i,j}$y​i,j​ pass a interpolation to restore input image resolution. Then use a threshold(Smoothing Prior) to get the final segmentation results.

Log-Sum-Exp(LSE)

$s^k = \frac{1}{r}\log \left[ \frac{1}{h^o w^o} \sum\limits_{i.j} exp\left( r s_{i,j}^k \right)\right]$sk=r1​log[howo1​i.j∑​exp(rsi,jk​)]
LSE is a pooling method for $Y:(|C|+1) \times h^{o} \times w^{o}$Y:(∣C∣+1)×ho×wo to $s:(|C|+1) \times 1 \times 1$s:(∣C∣+1)×1×1, it is more smooth. When $s$s is high LSE similar to max pooling, $r$r low LSE similar to average pooling.

For accuracy, performance be more high compare to max pooling and sum pooling.

Summary

• LSE is smooth pooling than max and average pooling. Maybe it is useful.