Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Recombinator Networks: Learning Coarse-to-Fine Feature Aggregation

About

Deep neural networks with alternating convolutional, max-pooling and decimation layers are widely used in state of the art architectures for computer vision. Max-pooling purposefully discards precise spatial information in order to create features that are more robust, and typically organized as lower resolution spatial feature maps. On some tasks, such as whole-image classification, max-pooling derived features are well suited; however, for tasks requiring precise localization, such as pixel level prediction and segmentation, max-pooling destroys exactly the information required to perform well. Precise localization may be preserved by shallow convnets without pooling but at the expense of robustness. Can we have our max-pooled multi-layered cake and eat it too? Several papers have proposed summation and concatenation based methods for combining upsampled coarse, abstract features with finer features to produce robust pixel level predictions. Here we introduce another model --- dubbed Recombinator Networks --- where coarse features inform finer features early in their formation such that finer features can make use of several layers of computation in deciding how to use coarse features. The model is trained once, end-to-end and performs better than summation-based architectures, reducing the error from the previous state of the art on two facial keypoint datasets, AFW and AFLW, by 30\% and beating the current state-of-the-art on 300W without using extra data. We improve performance even further by adding a denoising prediction model based on a novel convnet formulation.

Sina Honari, Jason Yosinski, Pascal Vincent, Christopher Pal• 2015

Related benchmarks

TaskDatasetResultRank
Facial Landmark Detection300-W (Common)
NME0.0467
180
Facial Landmark Detection300-W (Fullset)
Mean Error (%)5.41
174
Facial Landmark Detection300W (Challenging)
NME8.44
159
Facial Landmark DetectionAFLW Full
NME5.6
101
Face Alignment300W Fullset (test)--
82
Facial Landmark Detection300W
NME5.41
52
Facial Landmark Detection300-W Challenging Subset
Mean Error8.44
49
Facial Landmark Detection300-W public Challenging inter-pupil normalization (test)
NME9
46
Facial Landmark DetectionAFLW Front
NME5.36
38
Landmark LocalizationIBUG 300-W (test)
NME (%)8.44
31
Showing 10 of 23 rows

Other info

Follow for update