In this paper, we introduce a new regularization -- named Fishr -- that enforces domain invariance in the space of the gradients of the loss: specifically, the domain-level variances of gradients are matched across training domains.
Ranked #10 on Domain Generalization on Office-Home
Recent strategies achieved ensembling "for free" by fitting concurrently diverse subnetworks inside a single base network.
Ranked #6 on Image Classification on Tiny ImageNet Classification
The second stage combines a colorname-attention (dependent of the detected color) with an object-attention (dependent of the clothing category) and finally weights a spatial pooling over the image pixels' RGB values.
Similarly to self-training methods, the predictions of these initial detectors mitigate the missing annotations on the complementary datasets.